1: \documentstyle[12pt,aasms4,flushrt]{article}
2: %\documentstyle[11pt,aaspp4]{article}
3: %\input psfig.tex
4: \newcommand{\etal}{et al.\ }
5: \newcommand{\etalb}{et al.}
6: \newcommand{\Lya}{\mbox{Ly$\alpha$} }
7: \newcommand{\Ly}{\mbox{Ly$\alpha$}}
8:
9: \def\gsim{\;\rlap{\lower 2.5pt
10: \hbox{$\sim$}}\raise 1.5pt\hbox{$>$}\;}
11: \def\lsim{\;\rlap{\lower 2.5pt
12: \hbox{$\sim$}}\raise 1.5pt\hbox{$<$}\;}
13: \newcommand\beq{\begin{equation}}
14: \newcommand\eeq{\end{equation}}
15: \def\lya{Ly$\alpha$~}
16: \def\v{\vspace{-0.1in}}
17:
18: \begin{document}
19: \Large
20: \centerline{\bf The Formation of the First Low-Mass Stars From}
21: \centerline{\bf Gas With Low Carbon and Oxygen Abundances}
22: %\centerline{\bf Minimum Carbon and Oxygen Abundances Required}
23: %\centerline{\bf for the Formation of the First Low-Mass Stars}
24: \normalsize
25: \author{\bf Volker Bromm and Abraham Loeb}
26:
27: \noindent
28: Astronomy Department, Harvard University, 60 Garden Street,
29: Cambridge, MA 02138, USA
30:
31: \bigskip
32: \centerline{Appeared in {\it Nature}, {\bf 425}, 812--814 (2003)}
33: \medskip
34:
35: \vskip 0.2in
36: \hrule
37: \vskip 0.2in
38: \noindent
39: {\bf The first stars in the Universe are predicted to have been much more
40: massive than the Sun\cite{abn02}$^-$\cite{nu02}. Gravitational condensation
41: accompanied by cooling of the primordial gas due to molecular hydrogen,
42: yields a minimum fragmentation scale of a few hundred solar
43: masses. Numerical simulations indicate that once a gas clump acquires this
44: mass, it undergoes a slow, quasi-hydrostatic contraction without further
45: fragmentation\cite{abn02}$^,$\cite{bcl02}.
46: %In the absence of heavy elements, low-mass stars could not have formed.
47: Here we show that as soon as the primordial gas -- left over from the Big
48: Bang -- is enriched by supernovae to a carbon or oxygen abundance as small
49: as $\sim 0.01$--$0.1\%$ of that found in the Sun, cooling by singly-ionized
50: carbon or neutral oxygen can lead to the formation of low-mass stars. This
51: mechanism naturally accommodates the discovery\cite{HEStar} of solar mass
52: stars with unusually low ($10^{-5.3}$ of the solar value) iron abundance
53: but with a high ($10^{-1.3}$ solar) carbon abundance. The minimum stellar
54: mass at early epochs is partially regulated by the temperature of the
55: cosmic microwave background. The derived critical abundances can be used
56: to identify those metal-poor stars in our Milky Way galaxy with elemental
57: patterns imprinted by the first supernovae.}
58:
59: The microphysics of cooling due to molecular hydrogen (H$_{2}$) within the
60: first gas clouds to condense in the Universe, introduces a characteristic
61: temperature ($T_c \sim 100$--$200$~K) and hydrogen density ($n_c \sim
62: 10^{4}$~cm$^{-3}$) into the formation process of the first
63: stars\cite{abn02}$^,$\cite{bcl02}. The temperature floor is set by the
64: energy separation of the lowest-lying rotational levels of H$_{2}$, and the
65: characteristic density leads to the thermalization of these levels, at
66: which point cooling becomes less efficient. The minimum mass scale for
67: fragmentation is approximately given by the Jeans mass, $M_{J}\simeq 400
68: M_{\odot}(T/200{\rm\, K})^{3/2}(n/10^{4}{\rm cm}^{-3})^{-1/2}$. The final
69: mass of a star is determined by the accretion process onto the nascent
70: protostellar core, resulting in $M_{\ast}\simeq \alpha M_{J}$ with $\alpha
71: \la 0.5$ \cite{mkt02}$^,$\cite{cb03}. In contrast to the formation mode of
72: massive stars (population~III) at high redshifts, fragmentation is observed
73: to favor stars below a solar mass (population I and II) in the present-day
74: Universe. The transition between these fundamental modes is expected to be
75: mainly driven by the progressive enrichment of the cosmic gas with heavy
76: elements (or `metals'), which enable the gas to cool to lower temperatures.
77: The concept of a `critical metallicity', $Z_{\rm crit}$, has been
78: used\cite{om00}$^-$\cite{sch03} to characterize the transition between
79: population~III and population~II formation modes (where $Z$ denotes the
80: mass fraction contributed by all heavy elements). Previous
81: studies\cite{om00}$^,$\cite{bfcl01}$^,$\cite{sch03} have only constrained
82: this important parameter to within a few orders of magnitude, $Z_{\rm
83: crit}\sim 10^{-6}-10^{-3} Z_{\odot}$, under the implicit assumption of
84: solar relative abundances of metals. This assumption is likely to be
85: violated by the metal yields of the first supernovae (SNe) at
86: high-redshifts, for which strong deviations from solar abundance ratios are
87: predicted\cite{hw02}$^-$\cite{un03}. The cooling rate of the metals
88: depends on their ionization state, which in turn is controlled by the
89: ionizing backgrounds (UV and X-ray photons or cosmic rays) that were not
90: fully considered in earlier studies.
91:
92: Here we show that the transition between the above star formation modes is
93: driven primarily by fine-structure line cooling of singly-ionized carbon or
94: neutral atomic oxygen. We refine earlier estimates of $Z_{\rm crit}$ which
95: did not explicitly distinguish between different coolants, by introducing
96: separate critical abundances for carbon and oxygen, [C/H]$_{\rm crit}$ and
97: [O/H]$_{\rm crit}$, respectively, where [A/H]= $\log_{10}(N_{\rm A}/N_{\rm
98: H})-\log_{10}(N_{\rm A}/N_{\rm H})_{\odot}$, and a subscript `$\odot$'
99: denotes solar values. {\it Why are C and O identified as the key species
100: responsible for the global shift in the star formation mode?} We find that
101: under the temperature and density conditions that characterize
102: population~III star formation, the most important coolants are O\,I and
103: C\,II whose fine-structure lines dominate over all other metal
104: transitions\cite{hmk89}. Cooling due to molecules becomes important only at
105: lower temperatures, and cooling due to dust grains only at higher
106: densities\cite{hmk89}$^,$\cite{om00}$^,$\cite{sch03}.
107:
108: The conditions that terminate the predominance of massive, population~III
109: stars, have dramatic implications for the reionization history of the
110: Universe. Massive stars are much more efficient at producing ionizing
111: radiation than low mass stars, due to their high surface temperature ($\sim
112: 10^5$\,K). In particular, metal-free stars with a mass $\ga 100M_\odot$
113: produce $\sim 5\times 10^4$ hydrogen ionizing photons per baryon
114: incorporated into them, a yield higher by an order of magnitude than
115: obtained for a present-day mass function of
116: stars\cite{ts00}$^,$\cite{bkl01}. Massive stars can therefore reionize the
117: universe early\cite{cen03}$^-$\cite{wl03} at $z\sim 17$, as required by the
118: recent detection\cite{kog03} of large scale polarization anisotropies in
119: the cosmic microwave background (CMB). A strong UV flux just below the
120: Lyman limit ($h\nu < 13.6$\,eV) is predicted\cite{bl03} to be emitted by
121: the same stars that are responsible for the reionization of the IGM. This
122: soft UV radiation can penetrate the neutral IGM and ionize any trace amount
123: of neutral carbon due to its low first-ionization potential of
124: 11.26\,eV. Since carbon is highly underabundant, the UV background can
125: ionize carbon throughout the universe well before hydrogen reionization.
126: Oxygen, on the other hand, will be predominantly neutral prior to
127: reionization since its ionization potential is 13.62\,eV. Cooling is
128: mediated by excitations due to collisions of the respective metal with free
129: electrons or hydrogen atoms. At the low fractional abundances of electrons,
130: $x_{e}\la 10^{-4}$, expected in the neutral (rapidly recombining) gas at
131: $z\sim 15$, collisions with hydrogen atoms dominate. Even if the gas had
132: initially been ionized to a much higher degree due to the impact of the
133: supernova ejecta, it would have recombined to the level stated above within
134: its free-fall time [$t_{\rm ff}\sim 5\times 10^{5}
135: (n/10^{4}$\,cm$^{-3})^{-1/2}$\,yr]. This renders our analysis independent
136: of the very uncertain hydrogen-ionizing backgrounds at high redshifts
137: (cosmic rays or soft X-rays).
138:
139: We now derive the critical C and O abundances. Our starting point is the
140: characteristic state reached in primordial gas that cools only through
141: molecular hydrogen. For the gas to fragment further (instead of following a
142: slow contraction to a single massive star\cite{abn02}$^,$\cite{bcl02}),
143: additional cooling due to C\,II or O\,I is required. Fragmentation requires
144: that the radiative cooling rate be higher than the free-fall compressional
145: heating rate. We find the critical metal abundances by equating the two
146: rates: $\Lambda_{\rm C\,II, O\,I}(n,T)\simeq 1.5 n k_{\rm B} T/t_{\rm ff}$,
147: where $k_{\rm B}$ is Boltzmann's constant. In evaluating this threshold
148: condition, we solve the respective rate equations for C\,II (a two-level
149: system) and O\,I (a three-level system), including all possible radiative
150: and collisional transitions\cite{hmk89}. We assume that the emission of the
151: fine-structure lines proceeds under optically thin conditions, as
152: appropriate for the low atomic abundances considered here\cite{hmk89}. We
153: find the critical C and O abundances by setting $n\sim n_c$ and $T \sim
154: T_c$, and estimate the uncertainties by considering a plausible range near
155: these values ($100{\rm \,K}\la T\la 200{\rm \,K}$).
156:
157: In Fig.~1, we show the threshold C and O abundances that enable continued
158: fragmentation for various temperature and density values. It is convenient
159: to plot these quantities as functions of the Jeans mass ($\propto
160: T^{3/2}n^{-1/2}$). We find critical abundance values of [C/H]$_{\rm
161: crit}\simeq -3.5 \pm 0.1$ and [O/H]$_{\rm crit}\simeq -3.05 \pm 0.2$.
162: Strictly speaking, these threshold levels are the required abundances in
163: the gas phase. If a fraction of these elements were depleted onto dust
164: grains, then the total C and O elemental abundances would have to be
165: correspondingly higher. Considering the large current uncertainties in
166: modelling the properties and formation channels of dust at low metal
167: abundances\cite{sch03}, we assume the limiting case of zero C and O
168: depletion onto dust when comparing our theoretical predictions with
169: observed stellar abundances.
170:
171: We stress that even if sufficient C or O atoms are present to further cool
172: the gas, there will be a minimum attainable
173: temperature\cite{lar98}$^,$\cite{cb03} that is set by the interaction of
174: the atoms with the thermal CMB: $T_{\rm CMB}=2.7{\rm \,K}(1+z)$. At
175: $z\simeq 15$, this results in a minimum fragment mass of $M_{\ast}\sim 20
176: M_{\odot}(n_f/ 10^{4}{\rm \,cm}^{-3})^{-1/2}$, where $n_f> 10^{4}{\rm
177: \,cm}^{-3}$ is the density at which opacity prevents further
178: fragmentation\cite{rees76}. It is possible that the transition from the
179: high-mass to the low-mass star formation mode was modulated by the CMB
180: temperature and was therefore gradual\cite{cb03}, involving
181: intermediate-mass (`population~II.5') stars\cite{mbh03} at intermediate
182: redshifts. This transitional population can give rise to the faint SNe that
183: have been proposed to explain the observed abundance patterns in metal-poor
184: stars\cite{un02}$^,$\cite{un03}. These solar-mass stars themselves should
185: have formed out of metal-poor gas at relatively late times when the CMB
186: temperature was sufficiently low.
187:
188: In relating our results to theoretical nucleosynthetic yields, one has to
189: address the degree of mixing, or dilution, that the SN ejecta
190: experience. SNe at high redshifts occur in dark matter halos with shallow
191: potential wells, so that the resulting blast wave is predicted\cite{byh03}
192: to escape into the expanding IGM. The dilution mass can then be estimated
193: as $M_{\rm dil}\simeq 2 E_{\rm SN}/ v^2$, where $E_{\rm SN}$ is the
194: explosion kinetic energy, and $v=H(z)r$ is the Hubble velocity at a
195: distance $r$. Using standard cosmological parameters and ignoring radiative
196: losses\cite{byh03}, we get $M_{\rm dil}\simeq 10^{7} M_{\odot} (E_{\rm
197: SN}/10^{53}{\rm\, ergs})^{3/5}[(1+z)/16]^{-3/5}$. In the expected mass
198: range of the pair-instability supernovae\cite{hw02} (PISNe) and the
199: corresponding explosion energies $E_{\rm SN}\simeq 10^{51} -10^{53}$~ergs,
200: one finds dilution masses that are more than two orders of magnitude larger
201: than typical values for type~II SNe in nearby galaxies\cite{qw02}. However,
202: depending on the exact values of the ejected mass of C and O and the
203: respective explosion energies, a single SN event could already lead to
204: [C/H] and [O/H] ratios that `overshoot' the critical abundances. For
205: example, the abundances resulting from the explosion and subsequent mixing
206: of a PISN at the low-mass end\cite{hw02}, are [C/H]$\simeq -2.9 $ and
207: [O/H]$\simeq -2.4$, respectively, roughly $0.5$~dex above the critical
208: levels. Nevertheless, a single PISN at the high-mass end would produce final,
209: mixed abundances still below critical. Since the dilution mass scales as
210: the number of simultaneous SNe to the ${3/5}$ power while the ejecta mass
211: scales linearly with this number, multiple simultaneous SNe within the same
212: galaxy would be less effective at diluting their metal abundances into the
213: expanding IGM. But further dilution may occur during the hierarchical
214: assembly of galaxies, as infall may add metal-free gas from the IGM to the
215: enriched interstellar medium.
216:
217: In Fig.~2 we compare our theoretical thresholds to the observed C and O
218: abundances in metal-poor dwarf\cite{aker03} and giant\cite{cay03} stars in
219: the halo of our Galaxy. As can be seen, all data points lie above the
220: critical O abundance but a few cases lie below the critical C threshold.
221: All of these low mass stars are consistent with our model since the
222: corresponding O abundances lie above the predicted threshold. The
223: sub-critical [C/H] abundances could have either originated in the
224: progenitor cloud or from the mixing of CNO-processed material (with carbon
225: converted into nitrogen) into the stellar atmosphere during the red giant
226: phase. To guard against this a~posteriori processing of C, dwarf stars
227: should be preferred as they provide a more reliable record of the
228: primordial abundance pattern that was present at birth. However, since
229: dwarfs are fainter, another strategy would be to focus on giants which show
230: evidence for weak mixing (e.g. in terms of high $^{12}$C/$^{13}$C isotope
231: ratio or low $^{14}$N abundance). The current data on dwarf stars (filled
232: symbols in Fig. 2) does not yet reach C and O abundances that are
233: sufficiently low to probe the theoretical cooling thresholds. Note also
234: that the extremely iron-poor star HE0107-5240 has C and O abundances that
235: both lie above the respective critical levels (Bessell, M., et al., in
236: preparation). The formation of this low mass star ($\sim 0.8 M_{\odot}$) is
237: therefore consistent with the theoretical framework considered
238: here\cite{sch03}$^,$\cite{un03}.
239:
240: Our model suggests a powerful new probe of the first SNe to have exploded
241: in the Universe. By selecting an unbiased new sample of Galactic halo
242: stars with C or O abundances close to the theoretical critical values, it
243: should be possible to sample individual SN events originating in the
244: earliest epochs of cosmic star formation. The value of [C/Fe] would be
245: indicative of the particular SN type. For example, the very high value of
246: [C/Fe]$\simeq +4$ observed\cite{HEStar} in HE0107-5240, can be explained by
247: a core collapse SN which leaves behind a black hole after fallback of most
248: of the outer stellar envelope\cite{un03}. Modern surveys select metal-poor
249: stars based on their weak or absent Ca II K-lines. Our model suggests that a
250: better strategy to select truly `second generation' stars is to look for
251: stars with the lowest abundances of carbon and oxygen.
252:
253:
254:
255: \small
256:
257: \begin{thebibliography}{}
258:
259: \bibitem[$^1$]{abn02}\cite{abn02}Abel, T., Bryan, G. \& Norman, M.
260: The formation of the first star in the universe.
261: {\it Science} {\bf 295}, 93-98 (2002).
262: \v
263: \bibitem[$^2$]{bcl02}\cite{bcl02}Bromm, V., Coppi, P.S. \& Larson, R.B.
264: The formation of the first stars. I. The primordial star-forming cloud.
265: {\it Astrophys. J. } {\bf 564}, 23-51 (2002).
266: \v
267: \bibitem[$^3$]{nu02}\cite{nu02}Nakamura, F. \& Umemura, M.
268: The stellar initial mass function in primordial galaxies.
269: {\it Astrophys. J. } {\bf 569}, 549-557 (2002).
270: \v
271: \bibitem[$^4$]{HEStar}\cite{HEStar}Christlieb, N. \textit{et al.}
272: A stellar relic from the early Milky Way.
273: {\it Nature} {\bf 419}, 904-906 (2002).
274: \v
275: \bibitem[$^5$]{mkt02}\cite{mkt02}McKee, C. F. \& Tan, J. C.
276: Massive star formation in 100,000 years from turbulent and
277: pressurized molecular clouds.
278: {\it Nature } {\bf 416}, 59-61 (2002).
279: \v
280: \bibitem[$^6$]{cb03}\cite{cb03} Clarke, C. J. \& Bromm, V.
281: The characteristic stellar mass as a function of redshift.
282: {\it Mon. Not. R. Astron. Soc.} {\bf 343}, 1224-1230 (2003).
283: \v
284: \bibitem[$^{7}$]{om00}\cite{om00}Omukai, K.
285: Protostellar collapse with various metallicities.
286: {\it Astrophys. J. } {\bf 534}, 809-824 (2000).
287: \v
288: \bibitem[$^{8}$]{bfcl01}\cite{bfcl01}Bromm, V., Ferrara, A.,
289: Coppi, P.S. \& Larson, R.B.
290: The fragmentation of pre-enriched primordial objects.
291: {\it Mon. Not. R. Astron. Soc.} {\bf 328}, 969-976 (2001).
292: \v
293: \bibitem[$^{9}$]{sch02}\cite{sch02}Schneider, R., Ferrara, A.,
294: Natarajan, P. \& Omukai, K.
295: First stars, very massive black holes, and metals.
296: {\it Astrophys. J.} {\bf 571}, 30-39 (2002).
297: \v
298: \bibitem[$^{10}$]{mbh03}\cite{mbh03}Mackey, J., Bromm, V. \& Hernquist, L.
299: Three epochs of star formation in the high-redshift universe.
300: {\it Astrophys. J.} {\bf 586}, 1-11 (2003).
301: \v
302: \bibitem[$^{11}$]{sch03}\cite{sch03}Schneider, R., Ferrara, A.,
303: Salvaterra, R., Omukai, K. \& Bromm, V.
304: Low-mass relics of early star formation.
305: {\it Nature } {\bf 422}, 869-871 (2003).
306: \v
307: \bibitem[$^{12}$]{hw02}\cite{hw02}Heger, A. \& Woosley, S. E.
308: The nucleosynthetic signature of population III.
309: {\it Astrophys. J. } {\bf 567}, 532-543 (2002).
310: \v
311: \bibitem[$^{13}$]{qsw02}\cite{qsw02}Qian, Y.-Z., Sargent, W. L. W.
312: \& Wasserburg, G. J.
313: The prompt inventory from very massive stars and elemental abundances
314: in Ly$\alpha$ systems.
315: {\it Astrophys. J. } {\bf 569}, L61-L64 (2002).
316: \v
317: \bibitem[$^{14}$]{qw02}\cite{qw02}Qian, Y.-Z. \& Wasserburg, G. J.
318: Determination of nucleosynthetic yields of supernovae and very massive
319: stars from abundances in metal-poor stars.
320: {\it Astrophys. J. } {\bf 567}, 515-531 (2002).
321: \v
322: \bibitem[$^{15}$]{un02}\cite{un02}Umeda, H. \& Nomoto, K.
323: Nucleosynthesis of zinc and iron peak elements in population~III
324: type~II supernovae.
325: {\it Astrophys. J. } {\bf 565}, 385-404 (2002).
326: \v
327: \bibitem[$^{16}$]{sc03}\cite{sc03}Sneden, C. \& Cowan, J. J.
328: Genesis of the heaviest elements in the Milky Way Galaxy.
329: {\it Science } {\bf 299}, 70-75 (2003).
330: \v
331: \bibitem[$^{17}$]{un03}\cite{un03}Umeda, H. \& Nomoto, K.
332: First-generation black-hole-forming supernovae and the metal
333: abundance pattern of a very iron-poor star.
334: {\it Nature } {\bf 422}, 871-873 (2003).
335: \v
336: \bibitem[$^{18}$]{hmk89}\cite{hmk89}Hollenbach, D. \& McKee, C. F.
337: Molecule formation and infrared emission in fast interstellar shocks.
338: {\it Astrophys. J. } {\bf 342}, 306-336 (1989).
339: \v
340: \bibitem[$^{19}$]{ts00}\cite{ts00}Tumlinson, J. \& Shull, J. M.
341: Zero-metallicity stars and the effects of the first stars on reionization.
342: {\it Astrophys. J. } {\bf 528}, L65-L68 (2000).
343: \v
344: \bibitem[$^{20}$]{bkl01}\cite{bkl01}Bromm, V., Kudritzki, R.P. \& Loeb, A.
345: Generic spectrum and ionization efficiency of a heavy initial mass function
346: for the first stars.
347: {\it Astrophys. J. } {\bf 552}, 464-472 (2001).
348: \v
349: \bibitem[$^{21}$]{cen03}\cite{cen03} Cen, R.
350: The implications of {\it Wilkinson Microwave Anisotropy Probe} observations for
351: population~III star formation processes.
352: {\it Astrophys. J. } {\bf 591}, L5-L8 (2003).
353: \v
354: \bibitem[$^{22}$]{sok03}\cite{sok03} Sokasian, A., Yoshida, N.,
355: Abel, T., Hernquist, L. \& Springel V.
356: Cosmic reionisation by stellar sources: Population~III stars.
357: {\it Mon. Not. R. Astron. Soc.} (submitted); preprint available at
358: $\langle$http://xxx.lanl.gov/astro-ph/0302213$\rangle$ (2003).
359: \v
360: \bibitem[$^{23}$]{wl03}\cite{wl03}Wyithe, J. S. B. \& Loeb, A.
361: Was the universe reionized by massive metal-free stars?
362: {\it Astrophys. J. } {\bf 588}, L69-L72 (2003).
363: \v
364: \bibitem[$^{24}$]{kog03}\cite{kog03} Kogut, A., \textit{et al.}
365: Wilkinson Microwave Anisotropy Probe {\it WMAP} first year observations:
366: TE polarization.
367: {\it Astrophys. J.} (in the press); preprint available at
368: $\langle$http://xxx.lanl.gov/astro-ph/0302213$\rangle$ (2003).
369: \v
370: \bibitem[$^{25}$]{bl03}\cite{bl03} Bromm, V. \& Loeb, A.
371: Formation of the first supermassive black holes.
372: {\it Astrophys. J.} (in the press); preprint available at
373: $\langle$http://xxx.lanl.gov/astro-ph/0212400$\rangle$ (2003).
374: \v
375: \bibitem[$^{26}$]{lar98}\cite{lar98}Larson, R.B.
376: Early star formation and the evolution of the stellar initial
377: mass function in galaxies.
378: {\it Mon. Not. R. Astron. Soc.} {\bf 301}, 569-581 (1998).
379: \v
380: \bibitem[$^{27}$]{rees76}\cite{rees76} Rees, M. J.
381: Opacity-limited hierarchical fragmentation and the masses of protostars.
382: {\it Mon. Not. R. Astron. Soc.} {\bf 176}, 483-486 (1976).
383: \v
384: \bibitem[$^{28}$]{byh03}\cite{byh03} Bromm, V., Yoshida, N. \& Hernquist, L.
385: The first supernova explosions in the universe.
386: {\it Astrophys. J.} (submitted); preprint available at
387: $\langle$http://xxx.lanl.gov/astro-ph/0305333$\rangle$ (2003).
388: \v
389: \bibitem[$^{29}$]{aker03}\cite{aker03} Akerman, C.J., Carigi, L.,
390: Nissen, P.E., Pettini, M. \& Asplund, M.
391: The evolution of the C/O ratio in metal-poor halo stars.
392: {\it Astron. Astrophys.} (2003) (submitted) (2003).
393: \v
394: \bibitem[$^{30}$]{cay03}\cite{cay03} Cayrel, R. \textit{et al.}
395: Abundance patterns and supernova yields in the early Galaxy from C to Zn.
396: {\it Astron. Astrophys.} (2003) (submitted) (2003).
397:
398:
399: \end{thebibliography}
400:
401: \normalsize
402: \vskip 0.2in
403: \noindent
404: {\bf Acknowledgments} The authors thank Rennan Barkana and particularly Tim
405: Beers for helpful discussions, and are grateful to Tim Beers and Max
406: Pettini for making their data available prior to publication. This work
407: was supported in part by NSF, NASA, and the Guggenheim foundation (for
408: A.L.).
409:
410:
411: \vskip 0.2in
412: \noindent
413: {\bf Correspondence} and requests for materials should be addressed to
414: V.B. \\
415: (email: vbromm@cfa.harvard.edu) or A.L. (email: loeb@cfa.harvard.edu).
416:
417: \vskip 1in
418:
419: \begin{figure*}[hptb]
420: \caption{Required carbon and oxygen abundances (relative to solar values)
421: for cooling a clump of metal-poor gas faster than its free-fall
422: (compressional heating) time. The minimum abundances are shown as functions
423: of the initial Jeans mass of the clump (with higher abundances allowing the
424: clump to cool and further fragment). The upper panel shows the case of
425: singly ionized carbon, and the lower panel refers to neutral oxygen. The
426: curves are labeled by the hydrogen number density in units of cm$^{-3}$,
427: with the solid curve marking the characteristic density in pre-stellar
428: clumps of metal-poor gas, $n_c\sim 10^{4}$~cm$^{-3}$. Heavy filled dots
429: indicate particular temperature values. The critical abundance is
430: delineated by the horizontal dashed lines, with a heavy line indicating the
431: central value and the adjacent light lines bracketing the uncertainty
432: within the characteristic temperature range, $T_c\sim 100$--$200$~K, and
433: density, $\sim 10^{4}$~cm$^{-3}$, of pre-stellar clumps. The figure
434: indicates the minimum metal abundance at which fragmentation into low-mass
435: stars starts, but it does not provide information about the minimum mass of
436: the fragments at the end of the cooling process.}
437: \end{figure*}
438:
439: \begin{figure*}[hptb]
440: \plotone{figure1.ps}
441: \end{figure*}
442:
443:
444: \begin{figure*}[hptb]
445: \caption{Observed abundances in low-metallicity Galactic halo stars. For
446: both carbon (upper panel) and oxygen (lower panel), filled symbols
447: correspond to samples of dwarf and subgiant stars\cite{aker03}, and open
448: squares to a sample of giant stars\cite{cay03}. Both data sets were
449: obtained at a high signal-to-noise ratio using the UVES spectrograph on the
450: VLT. The dashed lines indicate the predicted level of critical carbon and
451: oxygen abundances from Fig.~1. According to our theoretical model, all
452: low-mass dwarf stars ought to lie above the critical threshold level of at
453: least one of the two elements. Indeed all stars in the existing data sets
454: satisfy this condition. We highlight the location of the extremely
455: iron-poor giant star HE0107-5240 (marked by {\bf x}), whose carbon and
456: oxygen abundances are both above the critical levels\cite{HEStar}.}
457: \end{figure*}
458:
459: \begin{figure*}[hptb]
460: \plotone{figure2.ps}
461: \end{figure*}
462:
463:
464: \end{document}
465: