1: \documentclass[manuscript]{emulateapj}

2: 

3: \usepackage{psfig,graphics} 

4: 

5: \def\subsun{\mbox{$_{\odot}$}}

6: 

7: \shorttitle{Three Modes of Star Formation}

8: \shortauthors{Smith, Turk, Sigurdsson, O'Shea, \& Norman}

9: 

10: \begin{document}

11: 

12: \title{Three Modes of Metal-Enriched Star Formation in the Early Universe}

13: 

14: \shorttitle{Three Modes of Metal-Enriched Star Formation}

15: 

16: \author{Britton D. Smith\altaffilmark{1}}

17: \affil{525 Davey Laboratory,

18:   Department of Astronomy \& Astrophysics,

19:   The Pennsylvania State University,

20:   University Park, PA 16802}

21: \email{brittons@origins.colorado.edu}

22: 

23: \author{Matthew J. Turk}

24: \affil{Kavli Institute for Particle Astrophysics and Cosmology, 

25:   2575 Sand Hill Rd., Mail Stop 29, 

26:   Menlo Park, CA 94025}

27: \email{mturk@slac.stanford.edu}

28: 

29: \author{Steinn Sigurdsson}

30: \affil{525 Davey Laboratory, 

31:   Department of Astronomy \& Astrophysics, 

32:   The Pennsylvania State University, 

33:   University Park, PA 16802}

34: \email{steinn@astro.psu.edu}

35: 

36: \author{Brian W. O'Shea\altaffilmark{2}}

37: \affil{Department of Physics \& Astronomy,

38:   Michigan State University,

39:   East Lansing, MI 48824}

40: \email{oshea@msu.edu}

41: 

42: \and

43: 

44: \author{Michael L. Norman}

45: \affil{Center for Astrophysics and Space Sciences,

46:   University of California at San Diego, La Jolla, CA 92093}

47: \email{mlnorman@ucsd.edu}

48: 

49: \altaffiltext{1}{Center for Astrophysics \& Space Astronomy,

50:   Department of Astrophysical \& Planetary Sciences,

51:   University of Colorado, Boulder, CO, 80309}

52: \altaffiltext{2}{Theoretical Astrophysics Group,

53:   Los Alamos National Laboratory,

54:   Los Alamos, NM 87545}

55: 

56: \begin{abstract}

57: Simulations of the formation of Population III (Pop III) stars suggest that they 

58: were much more massive than the Pop II and Pop I stars observed today. This 

59: is due to the collapse dynamics of metal-free gas, which is regulated by the 

60: radiative cooling of molecular hydrogen. We study how the collapse 

61: of gas clouds is altered by the addition of metals to the star-forming 

62: environment by performing a series of simulations of 

63: pre-enriched star formation at various metallicities.  To make a clean 

64: comparison to metal-free star formation, we use initial conditions identical 

65: to a Pop III star formation simulation, with low ionization and 

66: no external radiation other than the cosmic microwave background (CMB).  

67: For metallicities below the critical metallicity, $Z_{cr}$, 

68: collapse proceeds similarly to the metal-free case, and only massive objects form.  

69: For metallicities well above $Z_{cr}$, efficient cooling rapidly lowers the gas 

70: temperature to the temperature of the CMB.  The gas is 

71: unable to radiatively cool below the CMB temperature, and becomes 

72: thermally stable.  For high metallicities, $Z \ga 10^{-2.5} Z\subsun$, 

73: this occurs early in the evolution of the gas cloud, when the 

74: density is still relatively low.  The resulting cloud-cores show little or 

75: no fragmentation, and would most likely form massive stars.  If the metallicity is not vastly 

76: above $Z_{cr}$, the cloud cools efficiently but does not reach the CMB temperature, and 

77: fragmentation into multiple objects occurs.  We conclude that there were three distinct modes of star 

78: formation at high redshift ($z \ga 4$): a `primordial' mode, producing massive stars 

79: (10s to 100s $M\subsun$) at very low metallicities ($Z \la 10^{-3.75} Z\subsun$); a CMB-regulated mode, 

80: producing moderate mass (10s of $M\subsun$) stars at high metallicites ($Z \ga 10^{-2.5} Z\subsun$ at 

81: redshift $z\sim$15-20); and a low-mass (a few $M\subsun$) mode existing between those two metallicities.  

82: As the universe ages and the CMB temperature decreases, the range of the low mass mode extends to higher 

83: metallicities, eventually becoming the only mode of star formation.

84: \end{abstract}

85: 

86: \keywords{stars: formation, cosmology}

87: 

88: \section{Introduction}

89: 

90: Understanding the nature of the first stars in the universe is, in principle, a very 

91: straightforward problem to solve.  The initial conditions are defined by well-constrained 

92: cosmological parameters (e.g., \citet{2007ApJS..170..377S,2008arXiv0803.0547K}) and accepted 

93: calculations of 

94: Big Bang Nucleosynthesis (e.g., \citet{1993ApJS...85..219S}).  The relative simplicity of 

95: the chemistry of primordial gas \citep{1997NewA....2..181A,1998A&A...335..403G}, along with 

96: powerful numerical methods, have made it possible to accurately characterize the formation 

97: process of primordial stars from the assembly of their dark matter halos through the end of 

98: the optically-thin regime of the collapsing protostar \citep{2002Sci...295...93A,2002ApJ...564...23B,

99: 2004NewA....9..353B,2006ApJ...652....6Y,2007ApJ...654...66O,2007MNRAS.378..449G}.  Results 

100: from theory and numerical simulations suggest that the first stars were 10s to 

101: 100s $M\subsun$ \citep{2002Sci...295...93A,2003ApJ...589..677O,2004ApJ...603..383T,

102: 2006ApJ...652....6Y,2007ApJ...654...66O}.  

103: A direct calculation, however, involves simulating the complex processes of accretion and 

104: radiative feedback, a capability currently just beyond the state of the art.  

105: In contrast, the challenge of understanding the second generation of stars is significantly more 

106: complex.  The initial conditions of the first stars essentially depend only on the 

107: principles of cold dark matter cosmology and the properties of molecular hydrogen, 

108: but the initial conditions of the second stars 

109: require a complete solution to the formation and evolution of their predecessors as well as host of 

110: additional chemistry and physical processes.  The 

111: chemical composition and physical conditions of second-generation star-forming environments 

112: strongly depend on the exact masses of the first stars \citep{2002ApJ...567..532H,

113: 2005ASPC..336...79M,2006NuPhA.777..424N,2006astro.ph..8028R} and the mechanics of their 

114: supernovae \citep{2005ApJ...630..675K,2007ApJ...670....1G}.

115: 

116: At the heart of the problem regarding forming the second stars is understanding how the star 

117: formation 

118: process is altered by the introduction of the first metals in the universe, created in the 

119: explosions of the first stars.    

120: The chemical composition of a gas determines the efficiency with which it can cool radiatively.  

121: Cold, metal-free gas ($T < 10^{4}$ K) cools solely through ro-vibrational lines of H$_{2}$, whose 

122: lowest-lying transition has an energy equivalent temperature of $\sim$ 512 K, resulting in 

123: a minimum achievable temperature of $\sim$ 200 K.  The 

124: energy levels of H$_{2}$ become thermalized at rather low densities, $n$ $\sim$ 10$^{3-4}$ 

125: cm$^{-3}$, where $n$ is the number density, above which the cooling rate is only proportional to 

126: $n$, instead of $n^{2}$.  Numerical 

127: simulations have shown that this creates a stalling point in the star formation process, 

128: where the free-fall collapse of metal-free gas comes to a halt as the cooling time increases to  

129: be significantly above the dynamical time \citep{2002Sci...295...93A,2002ApJ...564...23B}.  Entry 

130: into this `loitering' phase marks the end of hierarchical fragmentation that occurs 

131: during the free-fall period, a fragmentation that occurs because the 

132: temperature is able to continually decrease with 

133: increasing density.  The final fragmentation mass scale is then set by the Jeans mass 

134: at the temperature and density corresponding to the point where the gas can no longer get colder 

135: with increasing density \citep{1985MNRAS.214..379L,2005MNRAS.359..211L}.  

136: In the metal-free case, this yields a mass scale of $\sim$ 1000 $M\subsun$, resulting in the 

137: high-mass nature of the first stars.

138: 

139: The addition of metals enhances the cooling rate at low temperatures through fine-structure 

140: and molecular transitions, as well as by continuum emission from dust grains.  At low to moderate 

141: densities ($n <$ 10$^{9}$ cm$^{-3}$), only metals in the gas phase contribute significantly to 

142: the cooling, unless the level of metal enrichment is very high \citep{2008MNRAS.385.1443S}.  

143: In the phases where H$_{2}$ cooling is strong ($n \la 10^{4}$ cm$^{-3}$, an increase in the cooling 

144: due to metals is first seen only at temperatures less 

145: than the 200 K temperature minimum that exists in metal-free star formation 

146: \citep{2008MNRAS.385.1443S}.  At abundances of $Z \la$ 10$^{-4}$ $Z\subsun$, where metal cooling is 

147: only significant for $T <$ 200 K, the low-density evolution of the cloud is identical 

148: to the metal-free case \citep{2000ApJ...534..809O,2001MNRAS.328..969B,2005ApJ...626..627O}.  

149: At higher metallicities, gas-phase metal cooling becomes strong enough for the collapsing cloud to 

150: bypass the loitering phase and to undergo continued fragmentation 

151: \citep{2001MNRAS.328..969B,2003Natur.425..812B,2006ApJ...643...26S,2007ApJ...661L...5S}.  

152: If dust grains are present, their influence becomes important at very high densities 

153: ($n >$ 10$^{12}$ cm$^{-3}$).  This has been shown to induce fragmentation for metallicities 

154: as low as 10$^{-5.5} Z\subsun$ \citep{2005ApJ...626..627O,2006MNRAS.369.1437S,2006ApJ...642L..61T,

155: 2008ApJ...672..757C}.  Calculations by \citet{2004MNRAS.351.1379S} predict that up to 30\% of the 

156: progenitor mass is converted into dust in a pair-instability supernova, but observations of Type II 

157: supernova in the local universe remnants have not returned conclusive evidence of dust 

158: \citep{2004MNRAS.355.1315G,2004Natur.432..596K}.  

159: In the context of star formation in the early universe, the arrival at 

160: this critical metallicity, $Z_{cr}$, at which fragmentation occurs beyond the capabilities of 

161: primordial gas, is predicted to be the point where the universal mode of star formation shifts 

162: permanently from the high-mass, solitary mode of the first stars, to the low-mass, 

163: multiply-producing mode that is presently observed  \citep{2003Natur.425..812B,2006ApJ...643...26S,

164: 2007ApJ...661L...5S}.

165: 

166: \begin{deluxetable*}{lccccccc}

167:   \tablecolumns{8}

168:   \tablewidth{0pt}

169:   \tablecaption{Simulations}

170:   \tablehead{

171:     \colhead{Run} &

172:     \colhead{IC} &

173:     \colhead{$Z$ (Z$\subsun$)} &

174:     \colhead{$z_{col}$} &

175:     \colhead{Grids} &

176:     \colhead{Cells ($\times10^{7}$)} &

177:     \colhead{$n_{max}$ (cm$^{-3}$)} &

178:     \colhead{$\Delta t_{col}$ (yr)}}

179:   \startdata

180:   r1\_mf       & 1 & 0            & 14.761 & 13790 & 6.49 & 5.96 $\times 10^{11}$ & -\\

181:   r1\_Z-6      & 1 & 10$^{-6}$     & 14.762 & 13015 & 6.42 & 6.43 $\times 10^{11}$ & 2.9 $\times10^{4}$\\

182:   r1\_Z-5      & 1 & 10$^{-5}$     & 14.783 & 12889 & 6.43 & 6.74 $\times 10^{11}$ & 5.7 $\times10^{5}$\\

183:   r1\_Z-4.25   & 1 & 10$^{-4.25}$  & 14.809 & 12955 & 6.40 & 6.75 $\times 10^{11}$ & 1.2 $\times10^{6}$\\

184:   r1\_Z-4      & 1 & 10$^{-4}$     & 14.830 & 12964 & 6.38 & 6.22 $\times 10^{11}$ & 1.8 $\times10^{6}$\\

185:   r1\_Z-3.75   & 1 & 10$^{-3.75}$  & 14.848 & 13003 & 6.38 & 6.97 $\times 10^{11}$ & 2.2 $\times10^{6}$\\

186:   r1\_Z-3.5    & 1 & 10$^{-3.5}$   & 14.874 & 12883 & 6.40 & 6.89 $\times 10^{11}$ & 2.9 $\times10^{6}$\\

187:   r1\_Z-3.25   & 1 & 10$^{-3.25}$  & 14.936 & 12888 & 6.36 & 3.97 $\times 10^{11}$ & 4.5 $\times10^{4}$\\

188:   r1\_Z-3      & 1 & 10$^{-3}$     & 15.073 & 12581 & 6.25 & 4.42 $\times 10^{11}$ & 7.9 $\times10^{6}$\\

189:   r1\_Z-2.5    & 1 & 10$^{-2.5}$   & 16.180 & 11187 & 5.69 & 4.70 $\times 10^{11}$ & 3.3 $\times10^{7}$\\

190:   r1\_Z-2      & 1 & 10$^{-2}$     & 19.481 & 8693  & 4.57 & 1.56 $\times 10^{11}$ & 8.8 $\times10^{7}$\\

191:   \tableline

192:   r2\_mf       & 2 & 0            & 17.409 & 8684  & 4.76 & 7.97 $\times 10^{11}$ & -\\

193:   r2\_Z-4      & 2 & 10$^{-4}$     & 17.555 & 8509  & 4.74 & 8.39 $\times 10^{11}$ & 2.5 $\times10^{6}$\\

194:   r2\_Z-3.5    & 2 & 10$^{-3.5}$   & 17.654 & 8476  & 4.71 & 8.20 $\times 10^{11}$ & 4.2 $\times10^{6}$\\

195:   r2\_Z-3      & 2 & 10$^{-3}$     & 17.955 & 8408  & 4.67 & 7.30 $\times 10^{11}$ & 9.2 $\times10^{6}$\\

196:   r2\_Z-2.5    & 2 & 10$^{-2.5}$   & 18.537 & 8022  & 4.51 & 4.41 $\times 10^{11}$ & 1.8 $\times10^{7}$\\

197:   r2\_Z-2      & 2 & 10$^{-2}$     & 20.441 & 7194  & 4.09 & 1.92 $\times 10^{11}$ & 4.4 $\times10^{7}$\\

198:   \tableline

199:   r3\_mf       & 3 & 0            & 23.885 & 7771  & 4.25 & 1.63 $\times 10^{12}$ & -\\

200:   r3\_Z-4      & 3 & 10$^{-4}$    & 23.966 & 7722  & 4.23 & 1.65 $\times 10^{12}$ & 6.7 $\times10^{5}$\\

201:   r3\_Z-3.5    & 3 & 10$^{-3.5}$  & 24.122 & 7640  & 4.21 & 1.42 $\times 10^{12}$ & 1.9 $\times10^{6}$\\

202:   r3\_Z-3      & 3 & 10$^{-3}$    & 24.390 & 7366  & 4.17 & 1.29 $\times 10^{12}$ & 4.1 $\times10^{6}$\\

203:   r3\_Z-2.5    & 3 & 10$^{-2.5}$  & 24.732 & 7145  & 4.13 & 1.20 $\times 10^{12}$ & 6.7 $\times10^{6}$\\

204:   r3\_Z-2      & 3 & 10$^{-2}$    & 25.028 & 7083  & 4.09 & 6.53 $\times 10^{11}$ & 8.9 $\times10^{6}$\\

205:   r3\_Z-2noCMB & 3 & 10$^{-2}$    & 25.255 & 7424  & 4.29 & 6.02 $\times 10^{11}$ & 1.1 $\times10^{7}$\\

206: \enddata

207: \tablecomments{$z_{col}$ is the redshift at the onset of runaway collapse.  The total number of grid cells includes 

208: those that are covered by child grids at higher levels of refinement.  $n_{max}$ is the proper maximum baryon number 

209: density within the box.  $\Delta$t$_{col}$ is the time difference to runaway collapse from the metal-free case.}

210: \label{tab:sims}

211: \end{deluxetable*}

212: 

213: The protostellar collapse of metal-enriched gas clouds has been studied extensively with 

214: one-zone models coupled to large chemical networks \citep{2000ApJ...534..809O,

215: 2003Natur.422..869S,2005ApJ...626..627O,2006MNRAS.369.1437S}.  These studies have 

216: produced insight into the evolution of the density and temperature of collapsing gas clouds 

217: with finite metallicities, but one-zone models cannot speak to the actual process of fragmentation, 

218: which requires attention to complex cloud geometries that can only be given by fully  

219: three-dimensional hydrodynamic simulations \citep{2007astro.ph..1733L}.  The first of such 

220: simulations were carried out by \citet{2001MNRAS.328..969B}, who included cooling from 

221: C, N, O, Fe, S, and Si, but not H$_{2}$, finding that clouds with metallicities, $Z$ $\ge$ 

222: 10$^{-3}$ $Z\subsun$ are able to fragment to mass scales lower than in the metal-free 

223: case.  These simulations had a mass resolution of only 100 $M\subsun$ and were, therefore, 

224: unable to investigate the formation of solar-mass stars.  More recently, 

225: \citet{2007ApJ...661L...5S} have performed a series of high-resolution simulations of 

226: metal-enriched gas collapse that were able to follow the evolution of gas fragments to 

227: sub-solar mass scales.  They, too, find that gas with metallicities, $Z$ $\ge$ 10$^{-3}$ 

228: $Z\subsun$ will fragment into multiple clumps, while gas with $Z$ $\le$ 10$^{-4}$ $Z\subsun$ 

229: will produce only one object.  However, \citet{2008MNRAS.385.1443S} reported the 

230: existence of regions in density and temperature that are thermally unstable in gas 

231: with metallicities as low as 10$^{-4}$ $Z\subsun$.  Fragmentation is traditionally thought 

232: to happen when the cooling time is less than the dynamical time, as the gas is able to 

233: cool and form perturbations before they can be smoothed out by sound waves.  When this 

234: condition, referred to as the fragmentation criterion, is satisfied, fragmentation 

235: can also be aided by thermal instabilities \citep{1965ApJ...142..531F}, where a slight 

236: decrease in temperature or increase in density leads to a higher cooling rate, causing 

237: differences in temperature between pockets of gas to grow in a runaway fashion.  

238: In the adaptive mesh refinement 

239: simulations of \citet{2007ApJ...661L...5S}, grid refinement was performed based on baryon 

240: and dark matter overdensities and by ensuring that the Truelove criterion, $l_{J}$ $<$ $\Delta x$, 

241: where $l_{J}$ is the local Jeans length and $\Delta x$ is the grid cell size, was satisfied 

242: \citep{1997ApJ...489L.179T}.  However, refinement was not performed when the cooling time was less than 

243: the hydrodynamic time-step.  While this was not explicitly wrong, since the radiative cooling 

244: solver iterates with time-steps that are no larger than 10\% of the cooling time, it 

245: may have artificially suppressed the growth of thermal instabilities that could have formed extra 

246: fragments.

247: 

248: We rerun the simulations of \citet{2007ApJ...661L...5S} with an additional refinement criterion 

249: which ensures that the hydrodynamic time-step is always less than the cooling time on the finest 

250: level of resolution.  

251: In addition, we extend the series of simulations to include lower metallicities and to more carefully 

252: examine the metallicity range 

253: between 10$^{-4}$ $Z\subsun$ and 10$^{-3}$ $Z\subsun$.  We also run similiar simulations with 

254: two extra sets of initial conditions to confirm the robustness of the results.  

255: We describe the setup of our simulations and the improvements over \citet{2007ApJ...661L...5S} in 

256: \S\ref{sec:setup}.  In \S\ref{sec:results}, we calculate the gas-phase critical metallicity, evaluate 

257: the validity of our assumption of optical thinness, and present the results of the suite of simulations.  

258: In \S\ref{sec:discussion}, we discuss our results in the context of star formation at high redshift and 

259: make the case for an initial mass function (IMF) that evolves over cosmic time.  We also include a 

260: discussion of the caveats and limitations of this work.  Finally, we conclude with a brief summary of the 

261: main conclusions of this work in \S\ref{sec:conclusion}.

262: 

263: \section{Simulation Setup} \label{sec:setup}

264: 

265: We perform a series of 24 primordial star formation simulations using the Eulerian adaptive mesh 

266: refinement hydrodynamics + N-body code, Enzo \citep{1997WSAMRGMBryan,2004CWAMROShea}.  Excluding 

267: our three metal-free control runs, the gas in each simulation is homogeneously pre-enriched to some 

268: non-zero metallicity.  As in \citet{2007ApJ...661L...5S}, we confine the simulations to constant 

269: metallicities with solar abundance patterns, saving the more realistic, and far more complicated, 

270: simulations of true 

271: second-generation star-forming environments, with heterogeneous metal-mixing and nonsolar abundance 

272: patterns, for a future work.

273: 

274: The nature of the initial conditions for our simulations are identical to those used in 

275: \citet{2007ApJ...661L...5S}.  The simulation box has a comoving size of 300 $h^{-1}$ kpc, 

276: with 128$^{3}$ grid cells on the top grid and three nested subgrids, each refining by a 

277: factor of 2, for an effective top grid resolution of 1024$^{3}$ cells.  The cosmological 

278: parameters have the following values:  $\Omega_{M}$ = 0.3, $\Omega_{\Lambda}$ = 0.7, 

279: $\Omega_{B}$ = 0.04, and Hubble constant, $h$ = 0.7, in units of 100 km s$^{-1}$ Mpc$^{-1}$.  

280: The power spectrum of initial density fluctuations is given by \citet{1999ApJ...511....5E}, 

281: with $\sigma_{8}$ = 0.9 and $n$ = 1.  Refined grids are created during the simulations when 

282: the baryon (dark matter) density is 4 (8) times greater than the mean density at that level.  

283: The density threshold for refinement decreases at higher levels.  

284: The local Jeans length is resolved by a minimum of 16 grid cells at all times, exceeding the 

285: Truelove criterion \citep{1997ApJ...489L.179T} by a factor of four along each coordinate axis.  

286: In addition, grid refinement occurs 

287: whenever the cooling time drops below the integration time step of the hydrodynamic solver.  

288: This final refinement criterion was not used in \citet{2007ApJ...661L...5S}.  During the simulation, 

289: a grid cell is flagged for refinement if one or more of any of these criteria are met.

290: 

291: We perform three sets of simulations.  Qualitatively, the three sets are the same.  They each 

292: have the same cosmological parameters, box size, and resolution.  The only difference between 

293: them is that their initial conditions were created with three unique randomizations of the initial 

294: density and velocity perturbations.  Thus, they represent three different realizations of the same 

295: problem.  The first set of initial conditions is the one used by \citet{2007ApJ...661L...5S}.  

296: The second and third sets were initially used by \citet{2007ApJ...654...66O} and correspond to the 

297: runs named L0\_30A and L0\_30D in that work.  For each set of initial conditions, we perform a 

298: metal-free control run.  Excluding the control runs, we run 10 simulations using initial condtions 

299: Set 1, with metallicities ranging from 10$^{-6} Z\subsun$ to 10$^{-2} Z\subsun$, 5 simulations 

300: using Set 2, and 6 simulations using Set 3, with metallicities from 10$^{-4} Z\subsun$ to 

301: 10$^{-2} Z\subsun$ for Sets 2 and 3.  The final simulation in Set 3 has a metallicity of 

302: 10$^{-2} Z\subsun$, but unlike the others, excludes the effect of the CMB on the cooling of gas.  

303: 

304: We use the second implementation of the optically-thin metal cooling method of 

305: \citet{2008MNRAS.385.1443S}.  This methods uses tabulated cooling functions created with 

306: the photoionization software, Cloudy \citep{1998PASP..110..761F}, for all elements heavier 

307: than He, up to atomic number 30 (Zn).  A solar abundance pattern is used for the metals in 

308: all of the simulations.  Since we only follow the evolution of the collapsing 

309: clouds up to densities of $\sim$ 10$^{12}$ cm$^{-3}$, we neglect the cooling from dust and assume 

310: the presence of only gas-phase metals.  In \S\ref{sec:tau}, we calculate the optical depth 

311: in the collapsing cloud for the most important high density, gas-phase metal coolants and find 

312: that the optically thin assumption holds at all times during simulations.  

313: The H/He chemistry is followed explicitly during the 

314: simulation as in \citet{1997NewA....2..181A} and \citet{1997NewA....2..209A}, but including reactions that 

315: extend its validity to above 10$^{8}$ cm$^{-3}$.  See \citet{2008MNRAS.385.1443S} for a full discussion 

316: of this method.  We neglect cooling from HD, as \citet{2002ApJ...564...23B} found its contribution to 

317: be negligible in situations where the initial ionization is low, such as in this work.  

318: However, it has been pointed out by \citet{2008AIPC..990...25G} that when the initial ionization is high, 

319: the HD fraction can become enhanced to the point where HD cooling is important.  

320: Heating from H$_{2}$ formation is not included in 

321: the chemical network.  This may contribute a significant source of heat when H$_{2}$ formation via 

322: three-body reactions becomes important at $n \ga 10^{9}$ cm$^{-3}$ \citep{2005ApJ...626..627O}.  

323: However, we argue in \S\ref{sec:frag} that our results are robust despite this shortcoming.  

324: The metal cooling data was created with the Linux computer cluster Lion-xo which is 

325: operated by the High Performance Computing Group at The Pennsylvania State University.  

326: 

327: As in the precursor to this work \citep{2007ApJ...661L...5S}, we do not assume the existence of an 

328: ionizing UV background.  The Population III stars responsible for the enrichment of the gas in our 

329: simulations have died, and we assume any living Pop III stars are most likely too distant 

330: to contribute significant amounts of radiation.  A major difference between our simulations 

331: and those of \citet{2007ApJ...661L...5S} is the choice of H$_{2}$ cooling rates.  

332: We use the updated H$_{2}$ cooling rates of \citet{1998A&A...335..403G}, whereas 

333: \citet{2007ApJ...661L...5S} used the older rates of \citet{1984ApJ...280..465L}.  The 

334: rates of \citet{1998A&A...335..403G} are generally lower than those of 

335: \citet{1984ApJ...280..465L}, causing the gas to take longer to cool in the newer 

336: simulations.  As a result, the moment of runaway-collapse in our simulations 

337: is systematically delayed from those of \citet{2007ApJ...661L...5S}, when 

338: comparing the simulations in Set 1.

339: 

340: Each simulation is initialized at $z$ = 99 and runs until the point where at least one dense, 

341: prestellar core forms within a $\sim 5 \times$ 10$^{5}$ $M\subsun$ dark matter halo, located in 

342: the center of the simulation box.  The simulations are stopped when a maximum refinement 

343: of 24 levels below the top grid has been reached.  This corresponds to a maximum density of 

344: $\sim$ 10$^{11}$ cm$^{-3}$, or roughly 3 $\times$ 10$^{-13}$ g cm$^{-3}$.  

345: The simulations were run on DataStar, an IBM Power4 machine at the San Diego Supercomputing Center. 

346: A summary of the final state of each simulation is given in Table \ref{tab:sims}.

347: 

348: \section{Results} \label{sec:results}

349: 

350: \subsection{Critical Metallicities} \label{sec:Z_crit}

351: 

352: The gas phase critical metallicity, $Z_{cr}$, has been estimated analytically by calculating the 

353: chemical abundance required for the cooling time to equal the dynamical time at the stalling 

354: point for metal-free gas, $n$ $\sim$ 10$^{3-4}$ cm$^{-3}$ and $T$ $\sim$ 200 K.

355: \citet{2003Natur.425..812B} performed this exercise with C and O, but excluding cooling from H$_{2}$, and 

356: \citet{2006ApJ...643...26S} did so with C, O, Fe, Si, and including H$_{2}$.  

357: \citet{2006ApJ...643...26S} considered densities above and below the stalling 

358: point as well.  The general consensus from these studies is that $Z_{cr}$ $\approx$ 

359: 10$^{-3.5}$ $Z\subsun$ at $n$ = 10$^{4}$ cm$^{-3}$.  

360: 

361: We observe a small amount of variance in the values of the density and temperature at 

362: which the temperature minimum occurs in our three metal-free runs.  Since the exact value of 

363: the critical metallicity depends on the precise conditions at the temperature minimum, i.e., 

364: temperature, total density, and H$_{2}$ density, we calculate the critical metallicity for 

365: each of the three runs separately.  To do this, we define the cooling time in the following 

366: way:

367: \begin{equation} \label{eqn:tcool}

368: t_{cool} = \frac{nkT}{(\gamma-1) [\Lambda_{H_{2}}^{\prime} + \Lambda_{metals}^{\prime}]},

369: \end{equation}

370: where $k$ is Boltzmann's constant, $\gamma$ is 5/3, and $\Lambda_{H_{2}}^{\prime}$ and 

371: $\Lambda_{metals}^{\prime}$ are the cooling rates from H$_{2}$ and the metals in units of 

372: [erg s$^{-1}$ cm$^{-3}$].  We use $\Lambda^{\prime}$ to avoid confusion with the term, $\Lambda$, 

373: which is often expressed in units of [erg s$^{-1}$ cm$^{3}$].  We then set Equation \ref{eqn:tcool} 

374: equal to the dynamical time,

375: \begin{equation} \label{eqn:tdyn}

376: t_{dyn} = \sqrt{\frac{3\pi}{16G\rho}},

377: \end{equation}

378: where $G$ is the gravitational constant and $\rho$ is the mass density.  We calculate $Z_{cr}$ 

379: for the conditions at the temperature minimum in each metal-free simulation using the H$_{2}$ 

380: cooling rates of \citet{1998A&A...335..403G} and the metal cooling rates from 

381: \citet{2008MNRAS.385.1443S}.  The results are shown in Table \ref{tab:zcrit}.  The systematic shift toward lower 

382: values of Z$_{cr}$ compared to previous calculations is most likely due to the fact that the 

383: minimum temperatures are slightly higher than 200 K, where both the H$_{2}$ and metal cooling rates 

384: are higher.  In addition, cooling from H$_{2}$ was not included in the calculation of 

385: \citet{2003Natur.425..812B}, meaning that more cooling from the metals would have been required.  

386: There is a correlation 

387: between $Z_{cr}$ and the collapse redshift of the simulation, with the highest value of $Z_{cr}$ 

388: coming from the highest collapse redshift, despite the lack of such a correlation for $n$, 

389: $n_{H_{2}}$, or $T$.  However, there is a correlation between the H$_{2}$ fraction and collapse 

390: redshift, which has been observed in the Pop III simulations of \citet{2007ApJ...654...66O}.  They 

391: find that the higher H$_{2}$ fractions result from the generally warmer gas in halos that collapse 

392: at higher redshifts, which is simply a function of the linear dependence upon redshift of the virial 

393: temperature.  This may be the dominant factor, 

394: but with only 3 data points it is unclear whether the observed trend is even significant.  In addition, 

395: the artificial nature of our initial conditions may make this finding inapplicable to the real world.

396: 

397: We see a trend with metallicity and collapse redshift that is similar to what was reported 

398: by \citet{2007ApJ...661L...5S}, where simulations with higher metallicities reach the 

399: runaway collapse phase earlier.  We define $\Delta t_{col}$ as the difference in time to 

400: runaway collapse between a simulation with non-zero metallicity and the metal-free run with 

401: the same initial conditions.  An increase in metallicity by 0.5 dex results in an increase in 

402: $\Delta t_{col}$ by a factor of approximately 1.3 to 4.  Similarly, when compared within a set, 

403: the simulations collapsing later have a higher number of total grids and grid cells in their 

404: final output, since the lower-density envelope gas has had more time to evolve and reach higher 

405: densities.

406: 

407: \begin{deluxetable}{lcccc}

408:   \tablecolumns{5}

409:   \tablewidth{0pt}

410:   \tablecaption{Critical Metallicities}

411:   \tablehead{

412:     \colhead{Run} &

413:     \colhead{$n$ [cm$^{-3}$]} &

414:     \colhead{$n_{H_{2}}$ [cm$^{-3}$]} &

415:     \colhead{$T$} & 

416:     \colhead{log($Z_{cr}$/$Z\subsun$)}}

417:   \startdata

418:   r1\_mf & 6.89$\times10^{3}$ & 3.42 & 283 & -4.08\\

419:   r2\_mf & 3.64$\times10^{3}$ & 1.86 & 214 & -3.90\\

420:   r3\_mf & 1.19$\times10^{4}$ & 6.53 & 260 & -3.85\\

421:   \enddata

422:   \tablecomments{Critical metallicities calculated for each metal-free simulation.  $n$, $n_{H_{2}}$, and $T$ 

423:     are the proper number density, H$_{2}$ number density, and temperature at the temperature minimum where H$_{2}$ 

424:     becomes thermalized in each of the metal-free simulations.  The final column is the log of the metallicity 

425:     required to equate the cooling time to the dynamical time for the conditions listed.}

426:   \label{tab:zcrit}

427: \end{deluxetable}

428: 

429: \subsection{Radial Profiles} \label{sec:radial}

430: 

431: In Figure \ref{fig:proj_08}, we show projections of mass-weighted mean number density for the central 0.5 pc 

432: surrounding the point of maximum baryon density for the final output of all runs in Set 1.  For the runs with 

433: metallicities near or below $Z_{cr}$ (10$^{-4.08} Z\subsun$ for this set of simulations), the central cores 

434: appear quite round and show no clear signs of forming more than one object.  In the metallicity range from 

435: 10$^{-3.75} Z\subsun$ to 10$^{-3.25} Z\subsun$, the cores appear increasing asymmetric, with at least one 

436: additional density maximum present.  However, at metallicities at or above 10$^{-3} Z\subsun$, the cores 

437: return to a more spherical shape with only a single density maximum.

438: 

439: \begin{figure*}

440:   \plotone{f1.eps}

441:   \caption{Projections of mass-weighed mean number density for the final output of all runs in 

442:     Set 1.  Each projection 

443:     is centered on the location of maximum density in the simulation box and has a width of 0.5 

444:     pc proper.  The labels in each panel indicate the log of the metallicity with respect to 

445:     solar for that run.  The images were made with the YT analysis toolkit 

446:     \cite[\texttt{yt.enzotools.org}]{SciPyProceedings_46}.

447:   } \label{fig:proj_08}

448: \end{figure*}

449: 

450: \begin{figure*}

451:   \plotone{f2.eps}

452:   \caption{Spherically-averaged, mass-weighted quantities as a function of radius from 

453:   the point of maximum density for 7 of the 10 simulations in Set 1.  A: number density 

454:   normalized to an $r^{-2}$ power law; B: temperature; C: enclosed gas mass; D: 

455:   radial velocity.  In each panel, the metallicities are $Z$ = 0 (solid-black), $10^{-6} Z\subsun$ 

456:   (solid-red), $10^{-5} Z\subsun$ (yellow), $10^{-4} Z\subsun$ (green), $10^{-3.5} Z\subsun$ (blue), 

457:   $10^{-3} Z\subsun$ (purple), $10^{-2.5} Z\subsun$ (dashed-black), and $10^{-2} Z\subsun$ 

458:   (dashed-red).} \label{fig:radial_08}

459: \end{figure*}

460: 

461: In Figure \ref{fig:radial_08}, we plot spherically averaged, mass-weighted values of the density, 

462: temperature, enclosed mass, and radial velocity as a function of radius from the point of maximum 

463: density for all simulations in Set 1.  On large scales, the density profiles follow an r$^{-2.2}$ 

464: power law.  To highlight the difference in density between each run, we plot in panel A of 

465: Figure \ref{fig:radial_08} the value of ($n \times r^{2}$), instead of simply $n$.  We choose to 

466: scale the density by $r^{2}$ instead of $r^{2.2}$ because it is easier to glean the true density from 

467: the figure.  Over 

468: most of the plotted range, the run with the lowest metallicity has the highest density.  In the 

469: isothermal collapse model of \citet{1977ApJ...214..488S}, the accretion rate is proportional to 

470: the cube of the sound speed, or $T^{3/2}$.  Figure \ref{fig:radial_08}B shows that while 

471: isothermality does not really apply, there is a clear correlation between the temperature and 

472: density.  The runs with the highest metallicity, and subsequently the coldest gas, are the least 

473: dense.  In addition, within individual runs, an increase in the gas temperature is matched by an 

474: increase in the density.

475: 

476: The instantaneous accretion rate at a given radius is a function of 

477: the infall velocity at that position.  Although there are some exceptions, the correlation between 

478: temperature/sound speed and infall velocity, with higher temperatures/sound speeds corresponding to higher 

479: velocities, generally holds.  This was also found to be true by \citet{2007ApJ...654...66O} in their simulations of 

480: Pop III star formation.  We find that the inflow is roughly transonic throughout the entire density range, showing 

481: that even though the collapsing clouds are not isothermal spheres, their accretion rates are still largely 

482: regulated by the sound speed.

483: 

484: \subsection{Optical Depth} \label{sec:tau}

485: 

486: \begin{figure*}

487:   \plotone{f3.eps}

488:   \caption{Optical depth for two of the most important high density, gas-phase coolants, Fe\textsc{ii} [25.99 $\mu m$] 

489:     (left) and O\textsc{i} [63.18 $\mu m$] (right), as a function of density for runs in Set 1.  The colors are the same as 

490:     in Figure \ref{fig:radial_08}.} \label{fig:tau_n_08}

491: \end{figure*}

492: 

493: The fragmentation of collapsing gas depends sensitively on its thermal evolution, which is effectively a 

494: measure of the radiative cooling properties of the gas as a function of density and temperature.  Since our 

495: radiative cooling method, described above, assumes optical thinness, the results of our simulations are only 

496: valid where $\tau \ll 1$.  The optically thin assumption begins to break down at densities $n \ga 10^{10}$ 

497: cm$^{-3}$ \citep{2005ApJ...626..627O}.  The optical depth at a frequency $\nu$ is expressed as

498: \begin{equation} \label{eqn:tau}

499:   \tau_{\nu} = \int  \kappa_{\nu} \: \rho  \: d\ell,

500: \end{equation}

501: where $\kappa_{\nu}$ is the opacity, $\rho$ is the mass density, and $d\ell$ is the distance traveled by a 

502: photon.  At a given density, the cloud has a characteristic size $r(\rho)$, shown in Figure \ref{fig:radial_08}A.  

503: If a photon emitted from a region with density $\rho$ is able to travel a distance $\sim r(\rho)$ without being 

504: re-absorbed, then it will not alter the thermal evolution of the cloud, and the assumption of optical thinness 

505: holds.  Using the Cloudy software, we calculate values of the absorption coefficient, $\alpha_{\nu} 

506: \equiv \kappa_{\nu} \: \rho$, as a function of density, metallicity, temperature, and frequency.  Equation 

507: \ref{eqn:tau} then takes the form

508: \begin{equation} \label{eqn:tau_sim}

509:   \tau_{\nu} = \alpha_{\nu}(\rho,Z,T) \: r(\rho).

510: \end{equation}

511: For metallicities $Z \la 10^{-2} Z\subsun$ and in the absence of dust, metal cooling is dominated by fine-structure 

512: transitions of O\textsc{i} and Fe\textsc{ii} \citep{2006ApJ...643...26S,2008MNRAS.385.1443S}.  In Figure 

513: \ref{fig:tau_n_08}, we plot the optical depth from Equation \ref{eqn:tau_sim} for the Fe\textsc{ii} line at 25.99 

514: $\mu$m and the O\textsc{i} line at 63.18 $\mu$m as a function of density using the 

515: spherically-averaged densities and temperatures shown in Figure \ref{fig:radial_08}.  

516: At low energies and in the absence of dust grains, the largest contributor to the opacity is free-free absorption, 

517: which is dominated by H for low metallicities.  As a result, the opacity in the energy range of interest is 

518: essentially independent of metallicity for metallicities less than solar.  When the metals become a considerable 

519: fraction of the total gas content, above $Z \sim 10 Z\subsun$ or so, the opacity at low energy for dust-free gas 

520: begins to grow with metallicity.  

521: In Figure \ref{fig:tau_n_08}, the apparent decrease in the optical depth at the 

522: highest metallicities is due to the flattening of the density profile at small radii.  As such, the 

523: final point for each curve in Figure \ref{fig:tau_n_08} should be ignored.  For values inside the point 

524: where the density profile flattens, it would be more reasonable to use the radius at which the density turns over 

525: for the calculation of $\tau$.  Fortunately, $\tau$ remains significantly less than 1 throughout all of our 

526: simulations, peaking at roughly 0.25 at the highest densities for the runs with the lowest metallicities.  The 

527: lower optical depth in the higher metallicity runs is due to the smaller characteristic size of the core, as seen 

528: in Figure \ref{fig:radial_08}A.  It should also be noted that some molecules are missing from the 

529: Cloudy dust-free chemistry network that could contribute to the opacity, such as TiO.  As such, the optical 

530: depths calculated could be slightly higher for the high metallicity simulations.  However, we show in 

531: \S\ref{sec:frag} that fragmentation in our simulations occurs well before this becomes a concern.

532: 

533: \subsection{Fragmentation} \label{sec:frag}

534: 

535: \begin{figure}

536:   \plotone{f4.eps}

537:   \caption{The number of bound clumps found within a sphere of radius 5 pc, centered on the 

538:   point of maximum density, as a function of metallicity for all of the simulations performed.  

539:   The grey bar in the bottom panel denotes the run where the temperature floor created by the CMB 

540:   was not included.  In each panel, the vertical dashed line represents the estimated value of 

541:   $Z_{cr}$ for each set from Table \ref{tab:zcrit}.} \label{fig:clumps}

542: \end{figure}

543: 

544: In order to quantify the degree of fragmentation within each run, we employ a clump finding 

545: algorithm to search for bound clumps within 5 pc of the density maximum.  As in 

546: \citet{1998ApJ...495..821T}, we define a clump, there referred to as a fragment, as ``the mass 

547: contained between a local density maximum and the lowest isodensity surface surrounding only that 

548: maximum,'' to quote that work.  We begin by identifying all grid cells within a sphere of radius 

549: 5 pc, centered on the point of maximum density.  We then create density contours using all the 

550: cells within the sphere.  In density space, the first contour spans the entire range of density 

551: within the sphere, effectively creating one large contour.  The contour becomes the parent clump 

552: of all other clumps that will be found as the process continues.  On the second iteration, we 

553: create contours with the same maximum as before, but with the minimum increased by 1/4 dex.  If 

554: more than one contour exists, those groups of cells become child clumps of the group made by the 

555: previous iteration.  The process continues in a recursive fashion, creating groups of cells based 

556: on contours of increasing minimum density, identified only within the cells of the parent clump.  

557: The clump finding ends when minimum countour density has reached the constant maximum.  Effectively, 

558: we create a family tree of clumps, with the very first group as the trunk of the tree.  During 

559: the clump finding process, a child clump is only kept if it is gravitationally self-bound, or has 

560: children of its own that are bound.  When determining whether a clump is bound, we consider the 

561: thermal energy of the gas as well its kinetic energy with respect to the bulk center of motion.  

562: 

563: In Figure \ref{fig:clumps}, we plot a histogram of all the number of clumps found in each run within the 

564: 5 pc sphere, as a function of the metallicity of the run.  For this plot, we only include bound child clumps 

565: with no children of their own.  When enlarging the radius of the sphere from 5 pc to 10 pc, no additional 

566: clumps were found in any of the runs.  Figure \ref{fig:clumps} confirms what is seen in Figure 

567: \ref{fig:proj_08}.  In all runs with metallicities below $Z_{cr}$, only a single bound clump is 

568: found.  As the metallicity increases, the number of clumps increases, then decreases back to only 

569: a single clump for the highest metallicities, with the exception of r3\_Z-2, which has 2 bound clumps.  

570: The range of metallicities where fragmentation occurs 

571: is consistent between Sets 1 and 3, but offset by 0.5 dex toward higher metallicities for Set 2.  

572: It is not clear what causes this offset, but the qualitative trend of increasing and then decreasing 

573: number of clumps exists in all 3 sets.  It is also worthwhile to note that runs r1\_Z-4, 

574: r1\_Z-3.75, and r2\_Z-3.5, while slightly above $Z_{cr}$, do not show fragmentation.  

575: 

576: \begin{deluxetable}{lcc}

577:   \tablecolumns{2}

578:   \tablewidth{0pt}

579:   \tablecaption{Fragmentation Properties}

580:   \tablehead{

581:     \colhead{Run} &

582:     \colhead{Clump Mass [$M\subsun$]} &

583:     \colhead{$n_{frag}$ [cm$^{-3}$]}}

584:   \startdata

585:   r1\_Z-3.5    & 54.4, 4.98 & 1.55 $\times 10^{7}$ \\

586:   r1\_Z-3.25   & 22.1       & 1.55 $\times 10^{6}$ \\

587:                & 5.46, 1.40 & 8.73 $\times 10^{6}$ \\

588:   r2\_Z-3      & 6.26, 1.89 & 2.75 $\times 10^{9}$ \\

589:   r2\_Z-2.5    & 95.6, 27.7 & 2.76 $\times 10^{4}$ \\

590:   r3\_Z-3.5    & 54.4, 4.98 & 1.55 $\times 10^{5}$ \\

591:   r3\_Z-2      & 289, 105   & 1.55 $\times 10^{4}$ \\

592:   r3\_Z-2noCMB & 11.7       & 2.76 $\times 10^{4}$ \\

593:                & 4.38, 4.06 & 4.91 $\times 10^{5}$ \\

594:                & 32.6, 0.49 & 8.73 $\times 10^{5}$ \\

595:   \enddata

596:   \tablecomments{Number density at which fragmentation occurs within simulations where multiple bound clumps 

597:     are identified and the masses of the clumps associated with that episode of fragmentation.  The fragmentation 

598:     density is taken to be the minimum density within a clump, taken from the 

599:     earliest data output where that clump is identified.}

600:   \label{tab:frag}

601: \end{deluxetable}

602: 

603: During the simulations, the Enzo code creates a snapshot of the entire box each time the maximum level 

604: of refinement increases.  This provides us with multiple data outputs as the central density increases 

605: during runaway collapse.  We ran the clump finder on all data outputs created during this period, searching 

606: for the first data output in which multiple clumps are found.  Within this output, we take the minimum 

607: density within the clump to be the density at which fragmentation occurred.  In Table \ref{tab:frag}, 

608: we list the fragmentation densities for all simulations in which multiple clumps were found.  In runs with 

609: 2 clumps, the fragmentation density is the same for both clumps, since it represents the lowest density 

610: contour for which the two objects are separate.  In run r1\_Z-3.25, which has 3 clumps, the core initially 

611: fragmented into 2 clumps at $n = 1.55 \times 10^{6}$ cm$^{-3}$.  One data output later, one of those clumps 

612: fragmented again at $n = 8.73 \times 10^{6}$ cm$^{-3}$.  A similar thing occurred in run r3\_Z-2noCMB 

613: with 4 clumps forming initially and one of those fragmenting again later.

614: 

615: At densities of roughly 10$^{9}$ cm$^{-3}$, H$_{2}$ formation via three-body reactions begin to rapidly 

616: increase the H$_{2}$ fraction \citep{2002Sci...295...93A,2002ApJ...564...23B}.  In our simulations, we find 

617: the H$_{2}$ fraction to be roughly constant at $\sim 10^{-3}$ until the density reaches $\sim 10^{9}$ cm$^{-3}$.  

618: At densities of a few $\times$ 10$^{10}$ cm$^{-3}$, the H$_{2}$ fraction has passed 10\% and continues to rise.  

619: The onset of rapid H$_{2}$ formation occurs slightly earlier (within a factor of a few in density) in the runs 

620: with higher metallicity due to the $T^{-1}$ dependence of the three-body rate coefficients.  Our chemical network 

621: lacks heating from H$_{2}$ formation ($\sim$ 4.4 eV per reaction), which may significantly raise the temperature 

622: of the gas when three-body reactions become important \citep{2005ApJ...626..627O}.  We test this by calculating 

623: the ratio of the heating rate produced by H$_{2}$ formation to the total rate of cooling (without H$_{2}$ formation 

624: heating) for each of the simulations.  We use the rate coefficients $k_{4}$ (H + H + H $\rightarrow$ H$_{2}$ + H) 

625: and $k_{6}$ (H + H + H$_{2}$ $\rightarrow$ H$_{2}$ + H$_{2}$) from \citet{1983ApJ...271..632P} and assume an energy 

626: injection of 4.4 eV per reaction.  We also include a cooling term representing the reverse of the above two 

627: reactions, using coefficients $k_{5}$ and $k_{7}$ of \citet{1983ApJ...271..632P}.  For metallicities $Z \le 

628: 10^{-3.5} Z\subsun$, the ratio of the H$_{2}$ formation heating to the total cooling is 0.1 at $n \sim 10^{9}$ 

629: cm$^{-3}$ and 1 at $n \sim 5 \times 10^{9}$ cm$^{-3}$.  For $Z = 10^{-3} Z\subsun$, this ratio is 0.1 at 

630: $n \sim 9 \times 10^{7}$ cm$^{-3}$ and 1 at $n \sim 6 \times 10^{8}$ cm$^{-3}$.  For $Z \ge 10^{-2.5} Z\subsun$, 

631: this ratio is 0.1 at $n \sim 3 \times 10^{7}$ cm$^{-3}$ and 1 at $n \sim 10^{8}$ cm$^{-3}$.  For run r3\_Z-2noCMB, 

632: which has the CMB removed, this ratio is 0.1 at $n \sim 10^{7}$ cm$^{-3}$ and 1 at $n \sim 10^{8}$ cm$^{-3}$.  

633: For simulations with equivalent metallicities, the above ratio varies maximally by a factor of a few due to 

634: minor differences in the temperature at a given density.  In all cases but one, fragmentation occurs well before 

635: H$_{2}$ formation heating becomes important.  The lone exception is run r3\_Z-3, where fragmentation occurs at 

636: $n \sim 3 \times 10^{9}$ cm$^{-3}$ and H$_{2}$ formation heating matches the total cooling at $n \sim 10^{9}$ 

637: cm$^{-3}$.  However, the core in this simulation does not fragment until it has already begun to reheat due to 

638: adiabatic compression, so it is not clear what effect the inclusion of this missing heating term would have had.

639: 

640: \begin{figure*}

641:   \plotone{f5.eps}

642:   \caption{Mass-weighted, average temperature as a function of number density for all runs in 

643:     Set 1.  The colors are the same as in Figure \ref{fig:radial_08}, including the runs with 

644:     metallicities $Z$ = $10^{-4.25} Z\subsun$ (dashed-yellow), $10^{-3.75} Z\subsun$ (dashed-green), and 

645:     $10^{-3.25} Z\subsun$ (dashed-blue).  The thin, black, dashed lines indicate lines of constant 

646:     Jeans mass in $M\subsun$.  The horizontal, blue, dashed line denotes the temperature of the CMB at $z$ = 19, 

647:     the approximate redshift of collapse for runs r1\_Z-2.5 and r1\_Z-2.  The central cores in these two 

648:     runs were both able to cool to the temperature of the CMB.} \label{fig:T_n_08}

649: \end{figure*}

650: 

651: \begin{figure*}

652:   \plotone{f6.eps}

653:   \caption{Mass-weighted, average temperature as a function of number density for all the runs in 

654:   Set 2 (left) and Set 3 (right).  For both panels, the metallicities are $Z$ = 0 (solid-black), 

655:   $10^{-4} Z\subsun$ (green), $10^{-3.5} Z\subsun$ (blue), $10^{-3} Z\subsun$ (purple), 

656:   $10^{-2.5} Z\subsun$ (dashed-black), and $10^{-2} Z\subsun$ (dashed-red).  In the bottom panel, 

657:   the dashed-orange line corresponds to run r3\_Z-2\_noCMB, with $Z = 10^{-2} Z\subsun$, but 

658:   with the CMB temperature floor removed.  All curves in Figures \ref{fig:T_n_08} and \ref{fig:T_n_real} 

659:   with the same colors refer to simulations with the same metallicities.  The thin, black, dashed lines indicate 

660:   lines of constant Jeans mass in $M\subsun$.  The horizontal, blue, dashed lines denote the temperature of the CMB at 

661:   $z$ = 19 (left) and 25 (right).} \label{fig:T_n_real}

662: \end{figure*}

663: 

664: In Figures \ref{fig:T_n_08} and \ref{fig:T_n_real}, we plot the number density vs. gas temperature 

665: for the final output of each simulation.  Due to the self-similar nature of the collapse, Figures 

666: \ref{fig:T_n_08} and \ref{fig:T_n_real} can also be used to understand the evolution of the central 

667: core throughout the collapse.  In all the runs with $Z < Z_{cr}$, the cooling is too low to prevent 

668: the temperature from rising at the H$_{2}$ thermalization density, $n$ $\sim$ 10$^{4}$ cm$^{-3}$.  

669: Therefore, the minimum fragmentation mass for these runs, set by the Jeans mass as the temperature 

670: minimum, is well over 1,000 $M\subsun$, which is nearly equivalent to the total enclosed mass.  

671: Even though runs r1\_Z-4 and r1\_Z-3.75 are above $Z_{cr}$, the additional cooling provided by the 

672: metals is not sufficient to significantly lower the minimum fragmentation mass.  For runs 

673: r1\_Z-3.5 and r1\_Z-3.25, the more efficient cooling lowers the minimum fragmentation mass to just 

674: over 100 $M\subsun$, which is approximately a factor of a few lower than the total mass within 1 pc.  

675: 

676: For the runs with the highest metallicities, as in runs r1\_Z-2.5 and r1\_Z-2, the gas cools 

677: all the way to the temperature of the CMB.  The cooling proceeds so efficiently that the gas has 

678: not had sufficient time to reach high densities before hitting the temperature floor of the CMB.  

679: Fragmentation can only continue as long as the temperature decreases with increasing density 

680: \citep{1985MNRAS.214..379L,2005MNRAS.359..211L}.  Although the temperature decreases slightly 

681: in runs r1\_Z-2.5 and r1\_Z-2 for densities greater than 10$^{3}$ cm$^{-3}$, the temperature 

682: minimum is effectively at $n$ = 10$^{3}$ cm$^{-3}$, where the gas reaches the CMB temperature.  

683: Near the CMB temperature, the value of the cooling rate, $\Lambda$, effectively becomes 

684: ($\Lambda(T) - \Lambda(T_{CMB})$).  Therefore, when the gas reaches the CMB temperature, the cooling 

685: rate drops to zero, and the cooling time becomes infinite.  The gas cloud becomes extremely 

686: thermally stable, preventing further fragmentation.

687: 

688: To verify that the CMB is indeed suppressing 

689: fragmentation, we run one simulation, r3\_Z-2\_noCMB, with the CMB temperature floor removed.  We 

690: choose initial conditions Set 3 for this exercise since it has the highest CMB temperature at the 

691: redshift of collapse and 

692: should therefore show the greatest contrast with the CMB removed.  In Figure \ref{fig:proj_cmb}, 

693: we show mass-weighted mean number density projections of the central 5 pc for runs, r3\_mf, r3\_Z-2, and 

694: r3\_Z-2\_noCMB.  Run r3\_Z-2 has a much clumpier structure than its metal-free counterpart, even with the CMB 

695: temperature floor present.  However, when the temperature floor is removed, the gas is able to 

696: collapse into a long, thin filament with far more small-scale structure.  As shown in Figure 

697: \ref{fig:clumps}, we find the most bound clumps in this run (4 within 1 pc of the density peak 

698: and 1 more within 5 pc).  In Figure \ref{fig:proj_cmb_temp}, 

699: we show mass-weighted mean temperature projections for the same runs as Figure \ref{fig:proj_cmb}.  We 

700: overlay contours of projected mean number density of 10$^{4}$ cm$^{-3}$.  Figure \ref{fig:T_n_real} shows that 

701: this is the approximate density at which the cloud in run r3\_Z-2 first reaches the CMB temperature.  

702: In run r3\_Z-2noCMB, it took $\sim$400,000 years for the central density to increase from 10$^{4}$ cm$^{-3}$ 

703: to 10$^{5}$ cm$^{-3}$, which is similar to the timescale in a dynamical collapse.  In run r3\_Z-2, the equivalent 

704: change in density took $\sim$1.9 million years, indicating that cooling to the CMB temperature has indeed 

705: ended free-fall collapse.  

706: 

707: In the projections of run r3\_Z-2 in 

708: Figure \ref{fig:proj_cmb_temp}, the 2 largest contours roughly represent the 2 bound clumps found within the 5 pc 

709: radius sphere.  The gas within the clumps has a very uniform temperature, as its cooling has been 

710: abruptly halted at the CMB temperature ($\sim$71 K).  The 2 clumps have masses of roughly 100 and 300 

711: $M\subsun$.  It is unlikely that they will fragment further, since the Jeans mass of each clump is nearly 

712: equivalent to its total mass.  At a temperature of 71 K, sound waves will travel $\sim$1.7 pc in the 1.9 Myr 

713: required for any additional fragments to increase in density by an order of magnitude once they have cooled to 

714: the CMB temperature.  Any additional fragments that might possibly condense out of lower densitiy gas must be 

715: approximately this large in order to collapse, suggesting they will also be very massive.  It is interesting 

716: to note that the two bound clumps in the simulation are also about this size.

717: 

718: In contrast, the gas inside the contours of run 

719: r3\_Z-2noCMB shows significantly more structure in temperature.  The cold knots in run r3\_Z-2noCMB seen in 

720: Figure \ref{fig:proj_cmb_temp} correspond to the high density regions seen in Figure \ref{fig:proj_cmb}.  4 of 

721: the 5 bound clumps are Jeans unstable, with masses of 4, 4, 12, and 33 $M\subsun$.  The fifth, with 

722: $M \sim 0.5 M\subsun$, is approximately 1/6 of its Jeans mass.  The most massive clump is about 5 times more 

723: massive than its Jeans mass.

724: 

725: \begin{figure*}

726:   \plotone{f7.eps}

727:   \caption{Projections of mass-weighted mean number density along the x (left), y (center), and z (right) axes for the 

728:     final output of runs r3\_mf with zero-metallicity (top), r3\_Z-2 with $Z$ = $10^{-2} Z\subsun$ (middle), and 

729:     r3\_-2noCMB with $Z$ = $10^{-2} Z\subsun$ and the CMB temperature floor removed (bottom).  Each projection 

730:     is centered on the location of maximum density in the simulation box and has a width of 5 

731:     pc proper.  The images were made with the YT analysis toolkit 

732:     \cite[\texttt{yt.enzotools.org}]{SciPyProceedings_46}.} \label{fig:proj_cmb}

733: \end{figure*}

734: 

735: \begin{figure*}

736:   \plotone{f8.eps}

737:   \caption{Projections of mass-weighted mean temperature density along the x (left), y (center), and z (right) axes 

738:     for the final output of runs r3\_mf with zero-metallicity (top), r3\_Z-2 with $Z$ = $10^{-2} Z\subsun$ (middle), 

739:     and r3\_-2noCMB with $Z$ = $10^{-2} Z\subsun$ and the CMB temperature floor removed (bottom).  Each projection 

740:     is centered on the location of maximum density in the simulation box and has a width of 5 pc proper.  The solid 

741:     lines show contours of projected mean number density of 10$^{4}$ cm$^{-3}$.  The images were made with the YT 

742:     analysis toolkit \cite[\texttt{yt.enzotools.org}]{SciPyProceedings_46}.} \label{fig:proj_cmb_temp}

743: \end{figure*}

744: 

745: \section{Discussion} \label{sec:discussion}

746: 

747: We have shown that fragmentation occurs within a collapsing cloud when the 

748: metallicity is above the critical metallicity.  

749: The exact value of the critical metallicity required to prevent an increase in temperature at the 

750: stalling point of H$_{2}$ varies slightly from halo to halo.  Within our three sets of initial 

751: conditions, the values of $Z_{cr}$ are correlated to the collapse redshift of the metal-free runs, 

752: with the highest Z$_{cr}$ corresponding to the highest redshift.  However, it is unclear whether 

753: this is significant.  If the metallicity is only marginally higher 

754: than $Z_{cr}$, fragmentation is unlikely to occur, as the increase in temperature at the H$_{2}$ 

755: stalling point is only delayed momentarily.  Thus, it is unlikely that a sharp transition in 

756: star formation mode occurs at just the moment when the critical metallicity is reached.

757: 

758: Fragmentation is suppressed when the metallicity is 

759: high enough such that the gas is able to cool to the temperature of the CMB when the central 

760: density is still relatively low.  

761: We confirm that the CMB is responsible for the observed 

762: suppression of fragmentation by running an identical simulation without the CMB.  In this simulation 

763: where the CMB is absent, we find more bound clumps than in any other of the runs in this study.  

764: 2 bound clumps were found in run r3\_Z-2, where the gas was able 

765: to cool to the CMB temperature.  However, both of these clumps were quite massive ($M \ga 100 M\subsun$), 

766: and we showed in \S\ref{sec:frag} that is is unlikely that they will fragment into smaller objects.  

767: We observe a small amount of variance in the metallicity range in which fragmentation occurs that 

768: does not appear to be related to the CMB.  Just as the exact value of $Z_{cr}$ seems to vary from 

769: halo to halo, we suspect that the range of metallicities where fragmentation occurs will also be 

770: influenced by the individual properties of a halo as well as its particular evolution.

771: 

772: The mass scale of collapsing clumps can be estimated from the Jeans mass at the end of the cooling 

773: phase.  This implies the existence of three distinct metallicity regimes for star formation.  In 

774: the first regime, $Z$ $\la$ $Z_{cr}$, which we refer to as the `primordial' mode, metals do not 

775: provide enough additional cooling to allow the gas 

776: temperature to continue to decrease monotonically with increasing density when the core reaches the 

777: H$_{2}$ thermalization density.  In this case, the collapse proceeds in a similar way to the 

778: metal-free scenario, resulting in the formation of a single, massive object.  There is also the 

779: potential for stars forming in this mode to be somewhat less massive than the very first stars.  

780: The fragmentation mass scale for extremely low metallicity gas ($Z \ll Z_{cr}$) may be lowered though 

781: compression by shocks from Pop III supernovae and enhanced cooling from HD in relic HII regions 

782: \citep{2003ApJ...586....1M,2006MNRAS.366..247J}.  

783: 

784: At the other extreme, we define 

785: $Z_{CMB}$ as the metallicity at which the gas can cool to the CMB temperature.  When $Z \gg Z_{CMB}$, 

786: the cloud-core will efficiently cool to the temperature of the CMB when the central 

787: density is still relatively low.  In this scenario, fragmentation is limited by cooling rapidly to 

788: the CMB temperature, as the mass scale is determined by the Jeans mass at the density when the core 

789: first reaches the CMB temperature.  We refer to this as the CMB-regulated star formation mode, similar 

790: to \citet{2007ApJ...664L..63T}.  As 

791: fragmentation is severely limited in this mode, these stars will most likely be more massive on average 

792: than the characteristic mass of stars forming today.

793: 

794: Finally, our simulations have shown that there 

795: exists a special range in metallicity, $Z_{cr} \le Z < Z_{CMB}$, where the core does not reheat 

796: at the metal-free stalling point, but also cannot cool all the way to the CMB temperature.  The minimum 

797: temperature is set only by the balance of radiative cooling and adiabatic heating.  The mass scale is 

798: not regulated externally by the CMB, but rather internally by the metallicity-dependent gas-cooling.  

799: Hence, we term this the metallicity-regulated star formation mode.  This mode produces the lowest mass 

800: stars of the three modes mentioned.

801: 

802: The CMB-regulated star formation mode creates a means by which a higher number of massive stars are 

803: formed in very early universe, when the CMB temperature was much higher.  As the universe evolves, the 

804: CMB temperature will slowly decrease, which will increase the metallicity required to reach the CMB 

805: temperature, referred to here as $Z_{CMB}$.  The decrease in the CMB temperature also means that the 

806: fragmentation mass scale will be lower at the point where the gas reaches the temperature floor.  Thus, 

807: the characteristic mass of stars produced by the CMB-regulated mode will slowly decrease with time.  

808: This behavior is in agreement with the model of an IMF that evolves with redshift formulated by 

809: \citet{1998MNRAS.301..569L}.  As the metallicity threshhold for the 

810: CMB-regulated mode advances to higher metallicity, the range of operation of the metallicity-regulated 

811: mode will extend to take its place.  We lack sufficient data in this study to predict the evolution of 

812: $Z_{CMB}$ with redshift.  However, in a paper to follow, we will map out the evolution of $Z_{CMB}$ vs. 

813: $z$ with additional simulations collapsing as much lower redshifts.  

814: Observations of nearby star-forming clouds show that the minimum 

815: achievable temperature in the local universe is roughly 10 K (e.g., \citet{1999ARA&A..37..311E}).  This 

816: implied that the CMB-regulated star formation mode is in operation up to $z \sim 2.7$, at the absolute 

817: latest.  

818: A growing amount of evidence has been presented that the stellar IMF evolves with redshift (e.g., 

819: \citet{2007MNRAS.379..985F,2007ApJ...664L..63T,2008ApJ...674...29V,2008MNRAS.385..147D,

820: 2008MNRAS.385..687W}).  Interestingly, 

821: \citet{2008ApJ...674...29V}, \citet{2008MNRAS.385..147D}, and \citet{2008MNRAS.385..687W} report 

822: evidence from high redshift that the 

823: IMF may deviate from the standard Salpeter IMF, favoring higher mass stars for $z \ga$ 2 - 4.

824: 

825: If dust is present in the very early universe, this would extend the range of the metallicity-regulated 

826: star formation mode to metallicities as low as $Z \sim 10^{-5.5} Z\subsun$ 

827: \citep{2005ApJ...626..627O,2006MNRAS.369.1437S,2006ApJ...642L..61T,2008ApJ...672..757C}.  The existence 

828: of low mass, hyper metal poor stars, HE0107-5240 \citep{2002Natur.419..904C}, and HE1327-2326 

829: \citep{2005Natur.434..871F}, both with [Fe/H] $<$ -5, may provide evidence of this.  Both of these stars 

830: show extremely enhanced C and O abundances, which would make their effective metallicities (in terms 

831: of the radiative cooling ability of gas with that abundance) much higher.  However, it is pointed out 

832: by \citet{2007ApJ...665.1361T,2007ApJ...664L..63T}, that the abundance patterns of these stars are best 

833: recreated by a scenario in which the C and O enhancement comes via binary mass transfer from an 

834: intermediate mass AGB star, meaning the stars are truly metal poor.  In that case, these two low-mass 

835: stars would likely require dust in order to form at such low metallicity.  \citet{2007ApJ...664L..63T} also 

836: shows that such a high fraction of carbon enhanced metal poor stars (CEMPs) requires a higher than normal 

837: fraction of more massive stars that go through the AGB phase.  

838: \citet{2007ApJ...664L..63T} claims that the evolution of CEMP fraction with metallicity (with a higher CEMP fraction 

839: at lower metallicity) already shows the influence of the CMB on the IMF.  If a CEMP star 

840: and its binary companion were formed from gas at the same metallicity, there would have to be process 

841: at work that would prevent the dust cooling fragmentation from forming only low mass stars.  The star 

842: formation models of \citet{2005ApJ...626..627O} do not indicate that the dust cooling phase that induces 

843: low mass fragmentation is able to reach the CMB temperature for $Z < 10^{-4} Z\subsun$.  This may simply 

844: imply that metal mixing from the first supernovae is higly heterogeneous, allowing stars to form 

845: simultaneously with largely different abundances.

846: 

847: In this work, we study only gas clouds with solar abundance patterns.  Most likely, the first metals in the 

848: universe will not have solar abundance patterns.  However, from the perspective of simulating metal-enriched 

849: star formation, it is not the specific elemental abundances of a gas cloud that are important, but rather the 

850: total cooling rate produced by the gas.  Therefore, given that it is the sum of the metals that is important, and 

851: not the abundance pattern, the results of this work are robust in spite of the fact that the abundance patterns used 

852: are likely to be incorrect.

853: 

854: There appears to be some discrepancy between our results and those of \citet{2007ApJ...660.1332J}, who see 

855: no evidence of fragmentation induced by gas-phase metal cooling.  This is potentially resolved by 

856: the fact that in that work, the gas collapse is only strictly followed up to densities of 

857: 5$\times10^{2}$ cm$^{-3}$ before sink particles are created.  Figures \ref{fig:T_n_08} and 

858: \ref{fig:T_n_real} show very little difference in the thermal structure of the gas for densities 

859: below 5$\times10^{2}$ cm$^{-3}$.  Additionally, our clump finding algorithm found only a single bound 

860: clump within every simulation when the analysis was performed on data outputs that were made when 

861: the maximum density was only $\sim10^{3}$ cm$^{-3}$.  Multiple clumps were only found when the clouds 

862: had reached somewhat higher densities.  Finally, the simulations in this work began with cold, neutral 

863: gas, whereas their simulations began with hot, ionized gas.  Had their simulations been run to higher 

864: densities, any dissimilarities might also be due to using different initial conditions.

865: 

866: \section{Conclusion} \label{sec:conclusion}

867: 

868: We have performed a series of high resolution simulations of metal-enriched star formation using 

869: cosmological, Pop III style initial conditions, and assuming fully homogeneous metal enrichment.  

870: We have shown that our results apply to more than a single star forming region by using 3 different 

871: sets of initial conditions with identical cosmological parameters and resolution, but with 3 unique 

872: random seeds with which to create the initial perturbations in the density and velocity fields.  From 

873: the results of these simulations, the main conclusions of this work are:

874: 

875: 1. Fragmentation does not occur when the metallicity is only slightly above $Z_{cr}$, since this only 

876: leads to a small delay in the onset of the loitering phase that is brought on by a decrease in the efficiency of 

877: H$_{2}$ cooling.  The density at which the temperature begins to increase with increasing density is only marginally 

878: higher than in the metal-free case, and therefore does not lead to a signicant lowering of the minimum Jeans 

879: mass.  Within our simulations, $Z_{cr}$ is roughly 10$^{-3.9} Z\subsun$.  

880: We find that fragmentation does not occur until the metallicity is roughly 0.5 dex above $Z_{cr}$.

881: 

882: 2. Fragmentation is suppressed when the metallicity is high enough such that the gas is able to cool to the 

883: temperature of the CMB when the density of the collapsing cloud is still relatively low ($n \sim 10^{4}$ cm$^{-3}$).  

884: The Jeans mass at the density and temperature at which the cloud first reaches the CMB temperature sets the minimum 

885: fragmentation mass within the cloud, and as such only massive clumps are able to form.

886: 

887: 3. Metal-enriched star formation occurs in three modes that are separated by two metallicity thresholds, $Z_{cr}$ 

888: and $Z_{CMB}$.  $Z_{cr}$ is the conventional critical metallicity and $Z_{CMB}$ is the metallicity at which the 

889: gas is able to cool to the temperature of the CMB at a given redshift.  At the approximate collapse redshifts 

890: of our simulations, $z \sim 20$, $Z_{CMB}$ is between 10$^{-3}$ and 10$^{-2.5} Z\subsun$.  The three modes of star 

891: formation are:

892: \begin{itemize}

893: \item The primordial mode ($Z < Z_{cr}$) - The additional cooling provided by the metals is not enough to 

894: significantly alter the thermal structure of the cloud relative to the metal-free case.  No fragmentation occurs 

895: and the star will have a mass similar to a Pop III star.  

896: \item The metallicity-regulated mode ($Z_{cr} < Z < Z_{CMB}$) - Metal cooling is high enough to allow the cloud to 

897: continue to cool past the metal-free loitering phase, but not high enough to cool it to the temperature of the CMB.  

898: The minimum fragmentation mass is set at lower temperatures and higher densities than in the primordial case and 

899: fragmentation into multiple objects occurs.  Based on the masses of clumps formed within our simulations, stars 

900: forming in this mode could have masses of the order of a few $M\subsun$ or less.

901: \item The CMB-regulated mode ($Z > Z_{CMB}$) - Fragmentation is suppressed when the cloud is able to cool to the CMB 

902: temperature, as is described in point 2 of the conclusions.  At minimum, stars forming in the CMB-regulated mode 

903: will be more massive than stars forming in the metallicity-regulated mode.  $Z_{CMB}$ will increase as the CMB 

904: temperature lowers with time.  As such, the masses of stars forming in the CMB-regulated mode will slowly decrease.  

905: This mode will vanish altogether when the CMB temperature reaches the observed minimum temperature of nearby 

906: molecular clouds ($T \sim$ 10 K at $z \sim$ 2.7).

907: \end{itemize}

908: 

909: 4. By demonstrating that the CMB can suppress fragmentation, we have provided a key conceptual piece 

910: CMB-IMF hypothesis \citep{2007ApJ...664L..63T}.  As pointed out by \citet{2007ApJ...664L..63T}, an 

911: IMF that evolves with redshift, producing more massive stars in the past, has consequences that may already 

912: be testable by observations.

913: 

914: \acknowledgments

915: We are extremely grateful to an anonymous referee who provided comments that significantly 

916: strengthened the arguments presented and helped to improve the manuscript a great deal.  

917: BDS would like to thank Jason Tumlinson and Simon Glover for extremely insightful discussions, as well as 

918: Gary Ferland and Peter van Hoof for their assistance with the Cloudy code.  

919: BDS and SS thank the Aspen Center for 

920: Physics for their hospitality during the 2008 Aspen Winter Conference On Astrophysics.  

921: This work was made possible by Hubble Space Telescope Theory Grant HST-AR-10978.01.  BWO and BDS carried 

922: out portions of this work under the auspices of the National Nuclear Security Administration of the U.S. 

923: Department of Energy at Los Alamos National Laboratory under Contract No. DE-AC52-06NA25396.  

924: BDS was supported at the University of Colorado, Boulder by NASA Theory grant NNX07AG77G.  BWO was 

925: supported by a LANL Director's Postdoctoral Fellowship (DOE LDRD grant 20051325PRD4).  MJT performed this 

926: work under the auspices of the Department of Energy at the Stanford Linear Accelerator Center in the Kavli 

927: Institute for Particle Astrophysics and Cosmology under Contract No. DE-AC02-76SF00515.  

928: The simulations were performed at SDSC with computing 

929: time provided by NRAC allocations MCA98N020 and TG-AST070010N.

930: 

931: \begin{thebibliography}{62}

932: \expandafter\ifx\csname natexlab\endcsname\relax\def\natexlab#1{#1}\fi

933: 

934: \bibitem[{{Abel} {et~al.}(1997){Abel}, {Anninos}, {Zhang}, \&

935:   {Norman}}]{1997NewA....2..181A}

936: {Abel}, T., {Anninos}, P., {Zhang}, Y., \& {Norman}, M.~L. 1997, New Astronomy,

937:   2, 181

938: 

939: \bibitem[{{Abel} {et~al.}(2002){Abel}, {Bryan}, \&

940:   {Norman}}]{2002Sci...295...93A}

941: {Abel}, T., {Bryan}, G.~L., \& {Norman}, M.~L. 2002, Science, 295, 93

942: 

943: \bibitem[{{Anninos} {et~al.}(1997){Anninos}, {Zhang}, {Abel}, \&

944:   {Norman}}]{1997NewA....2..209A}

945: {Anninos}, P., {Zhang}, Y., {Abel}, T., \& {Norman}, M.~L. 1997, New Astronomy,

946:   2, 209

947: 

948: \bibitem[{{Bromm} {et~al.}(2002){Bromm}, {Coppi}, \&

949:   {Larson}}]{2002ApJ...564...23B}

950: {Bromm}, V., {Coppi}, P.~S., \& {Larson}, R.~B. 2002, \apj, 564, 23

951: 

952: \bibitem[{{Bromm} {et~al.}(2001){Bromm}, {Ferrara}, {Coppi}, \&

953:   {Larson}}]{2001MNRAS.328..969B}

954: {Bromm}, V., {Ferrara}, A., {Coppi}, P.~S., \& {Larson}, R.~B. 2001, \mnras,

955:   328, 969

956: 

957: \bibitem[{{Bromm} \& {Loeb}(2003)}]{2003Natur.425..812B}

958: {Bromm}, V. \& {Loeb}, A. 2003, \nat, 425, 812

959: 

960: \bibitem[{{Bromm} \& {Loeb}(2004)}]{2004NewA....9..353B}

961: ---. 2004, New Astronomy, 9, 353

962: 

963: \bibitem[{{Bryan} \& {Norman}(1997)}]{1997WSAMRGMBryan}

964: {Bryan}, G. \& {Norman}, M.~L. 1997, in {Workshop on Structured Adaptive Mech

965:   Refinement Grid Methods}, ed. N.~{Chrisochoides}, {IMA Volumes in Mathematics

966:   No. 117} ({Springer-Verlag})

967: 

968: \bibitem[{{Christlieb} {et~al.}(2002){Christlieb}, {Bessell}, {Beers},

969:   {Gustafsson}, {Korn}, {Barklem}, {Karlsson}, {Mizuno-Wiedner}, \&

970:   {Rossi}}]{2002Natur.419..904C}

971: {Christlieb}, N., {Bessell}, M.~S., {Beers}, T.~C., {Gustafsson}, B., {Korn},

972:   A., {Barklem}, P.~S., {Karlsson}, T., {Mizuno-Wiedner}, M., \& {Rossi}, S.

973:   2002, \nat, 419, 904

974: 

975: \bibitem[{{Clark} {et~al.}(2008){Clark}, {Glover}, \&

976:   {Klessen}}]{2008ApJ...672..757C}

977: {Clark}, P.~C., {Glover}, S.~C.~O., \& {Klessen}, R.~S. 2008, \apj, 672, 757

978: 

979: \bibitem[{{Dav{\'e}}(2008)}]{2008MNRAS.385..147D}

980: {Dav{\'e}}, R. 2008, \mnras, 385, 147

981: 

982: \bibitem[{{Eisenstein} \& {Hu}(1999)}]{1999ApJ...511....5E}

983: {Eisenstein}, D.~J. \& {Hu}, W. 1999, \apj, 511, 5

984: 

985: \bibitem[{{Evans}(1999)}]{1999ARA&A..37..311E}

986: {Evans}, II, N.~J. 1999, \araa, 37, 311

987: 

988: \bibitem[{{Fardal} {et~al.}(2007){Fardal}, {Katz}, {Weinberg}, \&

989:   {Dav{\'e}}}]{2007MNRAS.379..985F}

990: {Fardal}, M.~A., {Katz}, N., {Weinberg}, D.~H., \& {Dav{\'e}}, R. 2007, \mnras,

991:   379, 985

992: 

993: \bibitem[{{Ferland} {et~al.}(1998){Ferland}, {Korista}, {Verner}, {Ferguson},

994:   {Kingdon}, \& {Verner}}]{1998PASP..110..761F}

995: {Ferland}, G.~J., {Korista}, K.~T., {Verner}, D.~A., {Ferguson}, J.~W.,

996:   {Kingdon}, J.~B., \& {Verner}, E.~M. 1998, \pasp, 110, 761

997: 

998: \bibitem[{{Field}(1965)}]{1965ApJ...142..531F}

999: {Field}, G.~B. 1965, \apj, 142, 531

1000: 

1001: \bibitem[{{Frebel} {et~al.}(2005){Frebel}, {Aoki}, {Christlieb}, {Ando},

1002:   {Asplund}, {Barklem}, {Beers}, {Eriksson}, {Fechner}, {Fujimoto}, {Honda},

1003:   {Kajino}, {Minezaki}, {Nomoto}, {Norris}, {Ryan}, {Takada-Hidai},

1004:   {Tsangarides}, \& {Yoshii}}]{2005Natur.434..871F}

1005: {Frebel}, A., {Aoki}, W., {Christlieb}, N., {Ando}, H., {Asplund}, M.,

1006:   {Barklem}, P.~S., {Beers}, T.~C., {Eriksson}, K., {Fechner}, C., {Fujimoto},

1007:   M.~Y., {Honda}, S., {Kajino}, T., {Minezaki}, T., {Nomoto}, K., {Norris},

1008:   J.~E., {Ryan}, S.~G., {Takada-Hidai}, M., {Tsangarides}, S., \& {Yoshii}, Y.

1009:   2005, \nat, 434, 871

1010: 

1011: \bibitem[{{Galli} \& {Palla}(1998)}]{1998A&A...335..403G}

1012: {Galli}, D. \& {Palla}, F. 1998, \aap, 335, 403

1013: 

1014: \bibitem[{{Gao} {et~al.}(2007){Gao}, {Yoshida}, {Abel}, {Frenk}, {Jenkins}, \&

1015:   {Springel}}]{2007MNRAS.378..449G}

1016: {Gao}, L., {Yoshida}, N., {Abel}, T., {Frenk}, C.~S., {Jenkins}, A., \&

1017:   {Springel}, V. 2007, \mnras, 378, 449

1018: 

1019: \bibitem[{{Glover}(2008)}]{2008AIPC..990...25G}

1020: {Glover}, S. 2008, in American Institute of Physics Conference Series, Vol.

1021:   990, First Stars III, ed. B.~W. {O'Shea} \& A.~{Heger}, 25--29

1022: 

1023: \bibitem[{{Green} {et~al.}(2004){Green}, {Tuffs}, \&

1024:   {Popescu}}]{2004MNRAS.355.1315G}

1025: {Green}, D.~A., {Tuffs}, R.~J., \& {Popescu}, C.~C. 2004, \mnras, 355, 1315

1026: 

1027: \bibitem[{{Greif} {et~al.}(2007){Greif}, {Johnson}, {Bromm}, \&

1028:   {Klessen}}]{2007ApJ...670....1G}

1029: {Greif}, T.~H., {Johnson}, J.~L., {Bromm}, V., \& {Klessen}, R.~S. 2007, \apj,

1030:   670, 1

1031: 

1032: \bibitem[{{Heger} \& {Woosley}(2002)}]{2002ApJ...567..532H}

1033: {Heger}, A. \& {Woosley}, S.~E. 2002, \apj, 567, 532

1034: 

1035: \bibitem[{{Jappsen} {et~al.}(2007){Jappsen}, {Glover}, {Klessen}, \& {Mac

1036:   Low}}]{2007ApJ...660.1332J}

1037: {Jappsen}, A.-K., {Glover}, S.~C.~O., {Klessen}, R.~S., \& {Mac Low}, M.-M.

1038:   2007, \apj, 660, 1332

1039: 

1040: \bibitem[{{Johnson} \& {Bromm}(2006)}]{2006MNRAS.366..247J}

1041: {Johnson}, J.~L. \& {Bromm}, V. 2006, \mnras, 366, 247

1042: 

1043: \bibitem[{{Kitayama} \& {Yoshida}(2005)}]{2005ApJ...630..675K}

1044: {Kitayama}, T. \& {Yoshida}, N. 2005, \apj, 630, 675

1045: 

1046: \bibitem[{{Komatsu} {et~al.}(2008){Komatsu}, {Dunkley}, {Nolta}, {Bennett},

1047:   {Gold}, {Hinshaw}, {Jarosik}, {Larson}, {Limon}, {Page}, {Spergel},

1048:   {Halpern}, {Hill}, {Kogut}, {Meyer}, {Tucker}, {Weiland}, {Wollack}, \&

1049:   {Wright}}]{2008arXiv0803.0547K}

1050: {Komatsu}, E., {Dunkley}, J., {Nolta}, M.~R., {Bennett}, C.~L., {Gold}, B.,

1051:   {Hinshaw}, G., {Jarosik}, N., {Larson}, D., {Limon}, M., {Page}, L.,

1052:   {Spergel}, D.~N., {Halpern}, M., {Hill}, R.~S., {Kogut}, A., {Meyer}, S.~S.,

1053:   {Tucker}, G.~S., {Weiland}, J.~L., {Wollack}, E., \& {Wright}, E.~L. 2008,

1054:   ArXiv e-prints, 803

1055: 

1056: \bibitem[{{Krause} {et~al.}(2004){Krause}, {Birkmann}, {Rieke}, {Lemke},

1057:   {Klaas}, {Hines}, \& {Gordon}}]{2004Natur.432..596K}

1058: {Krause}, O., {Birkmann}, S.~M., {Rieke}, G.~H., {Lemke}, D., {Klaas}, U.,

1059:   {Hines}, D.~C., \& {Gordon}, K.~D. 2004, \nat, 432, 596

1060: 

1061: \bibitem[{{Larson}(1985)}]{1985MNRAS.214..379L}

1062: {Larson}, R.~B. 1985, \mnras, 214, 379

1063: 

1064: \bibitem[{{Larson}(1998)}]{1998MNRAS.301..569L}

1065: ---. 1998, \mnras, 301, 569

1066: 

1067: \bibitem[{{Larson}(2005)}]{2005MNRAS.359..211L}

1068: ---. 2005, \mnras, 359, 211

1069: 

1070: \bibitem[{{Larson}(2007)}]{2007astro.ph..1733L}

1071: ---. 2007, ArXiv Astrophysics e-prints

1072: 

1073: \bibitem[{{Lepp} \& {Shull}(1984)}]{1984ApJ...280..465L}

1074: {Lepp}, S. \& {Shull}, J.~M. 1984, \apj, 280, 465

1075: 

1076: \bibitem[{{Mackey} {et~al.}(2003){Mackey}, {Bromm}, \&

1077:   {Hernquist}}]{2003ApJ...586....1M}

1078: {Mackey}, J., {Bromm}, V., \& {Hernquist}, L. 2003, \apj, 586, 1

1079: 

1080: \bibitem[{{Maeder} {et~al.}(2005){Maeder}, {Meynet}, \&

1081:   {Hirschi}}]{2005ASPC..336...79M}

1082: {Maeder}, A., {Meynet}, G., \& {Hirschi}, R. 2005, in Astronomical Society of

1083:   the Pacific Conference Series, Vol. 336, Cosmic Abundances as Records of

1084:   Stellar Evolution and Nucleosynthesis, ed. T.~G. {Barnes}, III \& F.~N.

1085:   {Bash}, 79--+

1086: 

1087: \bibitem[{{Nomoto} {et~al.}(2006){Nomoto}, {Tominaga}, {Umeda}, {Kobayashi}, \&

1088:   {Maeda}}]{2006NuPhA.777..424N}

1089: {Nomoto}, K., {Tominaga}, N., {Umeda}, H., {Kobayashi}, C., \& {Maeda}, K.

1090:   2006, Nuclear Physics A, 777, 424

1091: 

1092: \bibitem[{{Omukai}(2000)}]{2000ApJ...534..809O}

1093: {Omukai}, K. 2000, \apj, 534, 809

1094: 

1095: \bibitem[{{Omukai} \& {Palla}(2003)}]{2003ApJ...589..677O}

1096: {Omukai}, K. \& {Palla}, F. 2003, \apj, 589, 677

1097: 

1098: \bibitem[{{Omukai} {et~al.}(2005){Omukai}, {Tsuribe}, {Schneider}, \&

1099:   {Ferrara}}]{2005ApJ...626..627O}

1100: {Omukai}, K., {Tsuribe}, T., {Schneider}, R., \& {Ferrara}, A. 2005, \apj, 626,

1101:   627

1102: 

1103: \bibitem[{{O'Shea} {et~al.}(2004){O'Shea}, {Bryan}, {Bordner}, {Norman},

1104:   {Abel}, {Harknes}, \& {Kritsuk}}]{2004CWAMROShea}

1105: {O'Shea}, B.~W., {Bryan}, G., {Bordner}, J., {Norman}, M.~L., {Abel}, T.,

1106:   {Harknes}, R., \& {Kritsuk}, A. 2004, in {Lecture Notes in Computational

1107:   Science and Engineering}, Vol.~41, {Adaptive Mesh Refinement - Theory and

1108:   Applications}, ed. T.~{Plewa}, T.~{Linde}, \& V.~G. {Weirs}

1109: 

1110: \bibitem[{{O'Shea} \& {Norman}(2007)}]{2007ApJ...654...66O}

1111: {O'Shea}, B.~W. \& {Norman}, M.~L. 2007, \apj, 654, 66

1112: 

1113: \bibitem[{{Palla} {et~al.}(1983){Palla}, {Salpeter}, \&

1114:   {Stahler}}]{1983ApJ...271..632P}

1115: {Palla}, F., {Salpeter}, E.~E., \& {Stahler}, S.~W. 1983, \apj, 271, 632

1116: 

1117: \bibitem[{{Rockefeller} {et~al.}(2006){Rockefeller}, {Fryer}, \&

1118:   {Li}}]{2006astro.ph..8028R}

1119: {Rockefeller}, G., {Fryer}, C.~L., \& {Li}, H. 2006, ArXiv Astrophysics

1120:   e-prints

1121: 

1122: \bibitem[{{Santoro} \& {Shull}(2006)}]{2006ApJ...643...26S}

1123: {Santoro}, F. \& {Shull}, J.~M. 2006, \apj, 643, 26

1124: 

1125: \bibitem[{{Schneider} {et~al.}(2004){Schneider}, {Ferrara}, \&

1126:   {Salvaterra}}]{2004MNRAS.351.1379S}

1127: {Schneider}, R., {Ferrara}, A., \& {Salvaterra}, R. 2004, \mnras, 351, 1379

1128: 

1129: \bibitem[{{Schneider} {et~al.}(2003){Schneider}, {Ferrara}, {Salvaterra},

1130:   {Omukai}, \& {Bromm}}]{2003Natur.422..869S}

1131: {Schneider}, R., {Ferrara}, A., {Salvaterra}, R., {Omukai}, K., \& {Bromm}, V.

1132:   2003, \nat, 422, 869

1133: 

1134: \bibitem[{{Schneider} {et~al.}(2006){Schneider}, {Omukai}, {Inoue}, \&

1135:   {Ferrara}}]{2006MNRAS.369.1437S}

1136: {Schneider}, R., {Omukai}, K., {Inoue}, A.~K., \& {Ferrara}, A. 2006, \mnras,

1137:   369, 1437

1138: 

1139: \bibitem[{{Shu}(1977)}]{1977ApJ...214..488S}

1140: {Shu}, F.~H. 1977, \apj, 214, 488

1141: 

1142: \bibitem[{{Smith} {et~al.}(2008){Smith}, {Sigurdsson}, \&

1143:   {Abel}}]{2008MNRAS.385.1443S}

1144: {Smith}, B., {Sigurdsson}, S., \& {Abel}, T. 2008, \mnras, 385, 1443

1145: 

1146: \bibitem[{{Smith} \& {Sigurdsson}(2007)}]{2007ApJ...661L...5S}

1147: {Smith}, B.~D. \& {Sigurdsson}, S. 2007, \apjl, 661, L5

1148: 

1149: \bibitem[{{Smith} {et~al.}(1993){Smith}, {Kawano}, \&

1150:   {Malaney}}]{1993ApJS...85..219S}

1151: {Smith}, M.~S., {Kawano}, L.~H., \& {Malaney}, R.~A. 1993, \apjs, 85, 219

1152: 

1153: \bibitem[{{Spergel} {et~al.}(2007){Spergel}, {Bean}, {Dor{\'e}}, {Nolta},

1154:   {Bennett}, {Dunkley}, {Hinshaw}, {Jarosik}, {Komatsu}, {Page}, {Peiris},

1155:   {Verde}, {Halpern}, {Hill}, {Kogut}, {Limon}, {Meyer}, {Odegard}, {Tucker},

1156:   {Weiland}, {Wollack}, \& {Wright}}]{2007ApJS..170..377S}

1157: {Spergel}, D.~N., {Bean}, R., {Dor{\'e}}, O., {Nolta}, M.~R., {Bennett}, C.~L.,

1158:   {Dunkley}, J., {Hinshaw}, G., {Jarosik}, N., {Komatsu}, E., {Page}, L.,

1159:   {Peiris}, H.~V., {Verde}, L., {Halpern}, M., {Hill}, R.~S., {Kogut}, A.,

1160:   {Limon}, M., {Meyer}, S.~S., {Odegard}, N., {Tucker}, G.~S., {Weiland},

1161:   J.~L., {Wollack}, E., \& {Wright}, E.~L. 2007, \apjs, 170, 377

1162: 

1163: \bibitem[{{Tan} \& {McKee}(2004)}]{2004ApJ...603..383T}

1164: {Tan}, J.~C. \& {McKee}, C.~F. 2004, \apj, 603, 383

1165: 

1166: \bibitem[{{Truelove} {et~al.}(1997){Truelove}, {Klein}, {McKee}, {Holliman},

1167:   {Howell}, \& {Greenough}}]{1997ApJ...489L.179T}

1168: {Truelove}, J.~K., {Klein}, R.~I., {McKee}, C.~F., {Holliman}, II, J.~H.,

1169:   {Howell}, L.~H., \& {Greenough}, J.~A. 1997, \apjl, 489, L179+

1170: 

1171: \bibitem[{{Truelove} {et~al.}(1998){Truelove}, {Klein}, {McKee}, {Holliman},

1172:   {Howell}, {Greenough}, \& {Woods}}]{1998ApJ...495..821T}

1173: {Truelove}, J.~K., {Klein}, R.~I., {McKee}, C.~F., {Holliman}, II, J.~H.,

1174:   {Howell}, L.~H., {Greenough}, J.~A., \& {Woods}, D.~T. 1998, \apj, 495, 821

1175: 

1176: \bibitem[{{Tsuribe} \& {Omukai}(2006)}]{2006ApJ...642L..61T}

1177: {Tsuribe}, T. \& {Omukai}, K. 2006, \apjl, 642, L61

1178: 

1179: \bibitem[{{Tumlinson}(2007{\natexlab{a}})}]{2007ApJ...665.1361T}

1180: {Tumlinson}, J. 2007{\natexlab{a}}, \apj, 665, 1361

1181: 

1182: \bibitem[{{Tumlinson}(2007{\natexlab{b}})}]{2007ApJ...664L..63T}

1183: ---. 2007{\natexlab{b}}, \apjl, 664, L63

1184: 

1185: \bibitem[{Turk(2008)}]{SciPyProceedings_46}

1186: Turk, M. 2008, in Proceedings of the 7th Python in Science Conference, ed.

1187:   G.~Varoquaux, T.~Vaught, \& J.~Millman, Pasadena, CA USA, 46 -- 50

1188: 

1189: \bibitem[{{van Dokkum}(2008)}]{2008ApJ...674...29V}

1190: {van Dokkum}, P.~G. 2008, \apj, 674, 29

1191: 

1192: \bibitem[{{Wilkins} {et~al.}(2008){Wilkins}, {Trentham}, \&

1193:   {Hopkins}}]{2008MNRAS.385..687W}

1194: {Wilkins}, S.~M., {Trentham}, N., \& {Hopkins}, A.~M. 2008, \mnras, 385, 687

1195: 

1196: \bibitem[{{Yoshida} {et~al.}(2006){Yoshida}, {Omukai}, {Hernquist}, \&

1197:   {Abel}}]{2006ApJ...652....6Y}

1198: {Yoshida}, N., {Omukai}, K., {Hernquist}, L., \& {Abel}, T. 2006, \apj, 652, 6

1199: 

1200: \end{thebibliography}

1201: 

1202: \end{document}