Star formation and multi-phase interstellar medium in the ...sait.oat.ts.astro.it/MmSAI/85/PDF/347.pdf · Star formation and multi-phase interstellar medium in the rst galaxies

Mem. S.A.It. Vol. 85, 347c© SAIt 2014 Memorie della

Star formation and multi-phase interstellarmedium in the �rst galaxies

M. Ricotti1,2, O. Parry1, E. Polisensky1,3, and M. Bovill4

1 University of Maryland, Department of Astronomy, College Park, 20742 MD, USAe-mail: [email protected]

2 Sorbonne Universits, Institut Lagrange de Paris (ILP), 98 bis Bouldevard Arago 75014Paris, France

3 Naval Research Laboratory, Washington, D.C. 20375, USA4 Pontificia Universidad Catlica de Chile, Avda. Libertador Bernardo O’Higgins 340,

Santiago,Chile

Abstract. Star formation and metal enrichment in the first galaxies is discussed emphasiz-ing similarities to the properties of dwarf spheroidal galaxies in the Local Universe. I presentpreliminary results from new radiation-hydrodynamic cosmological simulations for the for-mation of the first galaxies performed with the ART code. The simulations include a detailedmodel for star formation in a multi-phase ISM, including H2 formation catalyzed by H− andon dust grains. The first metals are provided by Population III stars, while Population II starformation takes place in resolved molecular clouds. The properties of the first galaxies inthese new simulations are in agreement with previous lower resolution simulations in whichwas found remarkable similarities between the fossils of the first galaxies and the faintestdwarf spheroidal galaxies in the Local Group.

Key words. Dark ages, reionization, first stars – Stars: Population III – Cosmology: theory– ISM: general – Local Group – Galaxies:dwarf

1. Introduction

In cold dark matter cosmologies, small masshalos outnumber larger mass halos at any red-shift. However, the lower bound for the massof a galaxy is unknown, as are the typical lumi-nosity of the smallest galaxies and their num-bers in the universe. If an early population ofdwarf galaxies did form in significant num-ber, their relics should be found today in theLocal Group. These relics have been named“fossils of the first galaxies.” This talk is a re-view that summarizes ongoing efforts to simu-

late and identify these fossils around the MilkyWay and Andromeda.

Simulating the formation of the first galax-ies in a cosmologically representative volumeis complex because the results are sensitive tofeedback effects acting on cosmological scalesand the uncertain initial mass function (IMF)of the first stars. Progresses can be made onlyif we can constrain our models with observa-tions. HST has detected a few redshift z = 10candidates, but the bulk of the first popula-tion of dwarf galaxies is still undetected. Evenwith JWST we may not be able to probe thebulk of this population if, as simulations seem

348 Ricotti: Star formation in the first galaxies

Fig. 1. ((Left.) Global star formation rate (SFR) (solid lines) and mass in stars (dashed lines) as a functionof redshift in three simulations in which the sub-grid parameter that determines the star formation efficiencyin molecular clouds was varied by a factor of 10 above and below a fiducial value. The SFR is self-regulatedon a global scale. (Right.) Global SFR of population II stars (red solid line), population III stars (blue dashedline) and mass in stars (dotted black line) as a function of redshift.

Fig. 2. Half-light radii as a function of luminosityof the stellar spheroids of the first galaxies at z = 10in the simulations by Parry et al. (2014), in prepara-tion. The triangles refer to the radial extent of Pop IIIstars only and the squares to Pop II stars only. Theresult agrees with previous lower-resolution simula-tions (Ricotti et al. 2002a; Ricotti & Gnedin 2005).

to suggest, is intrinsically faint. However, apromising avenue to constrain models of thefirst galaxies consists in using near field ob-servations of the faintest dwarfs in the LocalGroup to identify the surviving fossils of thefirst galaxies (Ricotti & Gnedin 2005). Thediscovery of a population of ultra-faint dwarfssatellites of the Milky Way (Zucker et al. 2006;Belokurov et al. 2007; Majewski et al. 2007)with properties consistent with predictions ofsimulations of the fossils of the first galaxies

(Bovill & Ricotti 2009; Salvadori & Ferrara2009; Ricotti 2009, 2010), is an exciting obser-vational development warranting further stud-ies.

In this presentation I briefly review thetimeline of progresses in this field (§ 2.1), andthe main qualitative results from simulations ofthe formation of the first galaxies at z > 10(§ 2.2). I then touch on metal production andenrichment of the interstellar medium (ISM)and intergalactic medium (IGM) (§ 2.3) anddiscuss the evolution of fossil dwarfs from for-mation to redshift z = 0 and the consistencyof the models with observations of the MilkyWay satellites (§ 2.4). I conclude with a briefsummary (§ 3).

2. The first galaxies

In order to obtain sufficiently large densities toinitiate star formation in dark matter halos, gasneeds to cool by emitting radiation. In proto-galaxies with masses > 108 − 109 M�, thattypically form after the redshift of reioniza-tion (z < 10), the cooling is provided by hy-drogen Lyman-alpha emission, that is efficientat temperatures ∼ 20, 000 K, but is negligiblebelow T ∼ 10, 000 K. At temperature below10, 000 K, typical of gas in the first small masshalos, cooling can only be provided by metal

Ricotti: Star formation in the first galaxies 349

line cooling or by molecules. However, boththese cooling mechanisms may be absent inthe first galaxies. The first dwarf galaxies differwhen compared to present-day galaxies in twomain respects:

1. they lack important coolants – such as car-bon and oxygen – because the gas initiallyis of primordial composition, and

2. due to the small typical masses of the firstdark halos, the initial temperature of thegas is too low to cool by Lyman-alphaemission.

2.1. Timeline of progresses

A gas has primordial composition is unableto initiate star formation unless it can forma sufficient amount of primordial H2 via theH− chain. Tegmark et al. (1997) found thatan abundance xH2

>∼ 10−4 is required to initiatethe cooling and star formation. This require-ment implies that the first luminous halos havemasses > 105 M� and first appear at redshiftz ∼ 30.

Because molecular hydrogen is easily de-stroyed by far-ultraviolet (FUV) radiation inthe Lyman-Werner bands (11.3 eV < hν <13.6 eV) emitted by the first stars, it wasbelieved that the majority of galaxies withvvir < 20 km s−1 remain dark (e.g., Haimanet al. 2000). However, several studies have nowshown that even if the FUV radiation back-ground is strong, a small amount of H2 canform, particularly in relatively massive haloswith virial temperature of several thousandsof degrees (Wise & Abel 2007; O’Shea &Norman 2008). Thus, negative feedback fromFUV radiation only delays star formation in themost massive dwarfs (Machacek et al. 2001,2003). In a series of papers, Ricotti et al.(2002a,b, 2008) found that hydrogen ionizingradiation (hν > 13.6 eV) plays a far more im-portant role in regulating the formation of thefirst galaxies than FUV radiation. Thus, mod-els that do not include 3D radiative transfer ofhydrogen and Helium ionizing radiation can-not capture the most relevant feedback mech-anism that regulates galaxy formation in theearly universe.

Around the epoch of reionization the for-mation of dwarf galaxies with mass < 108 −109 M� (or circular velocity vvir

<∼ 20 km s−1),is inhibited by the increase in the Jeans massin the IGM. Thus, reionization feedback andnegative feedback by the FUV background de-termine the mass of the smallest galactic build-ing blocks. However, relatively recent episodes(at z < 1) of gas accretion and star forma-tion may take place in some fossils of the firstgalaxies that evolve in isolation due to theirlarge concentration (or equivalently, large massfor a given circular velocity) and the decreas-ing temperature of the intergalactic medium af-ter Helium reionization at z <∼ 3 (Ricotti et al.2000). Fossil dwarfs may be characterized ei-ther by a single old population of stars or by abimodal star formation history (Ricotti 2009).

2.2. Modern simulations

The first simulations of the the formation ofthe first galaxies in a cosmological volume in-cluding 3D radiation transfer were published

Circularity: angular

momentum

normalized to the

angular momentum of

a circular orbit with

the same energy

Parry, Ricotti, Gnedin, in preparation

Fig. 3. Circularity of population II stars (red his-tograms) and population III stars (blue histograms)in four first galaxies at z = 10 in the simulationsby Parry et al. (2014), in preparation. The circular-ity is defined as the angular momentum normalizedto the angular momentum of a circular orbit withthe same energy. The symmetry of the distributionaround zero indicates that the stars in the first galax-ies form as spheroids with little rotation.



SN ejecta

Cold phase

Warm phase

Molecular gas

Fig. 4. Multi-phase ISM in four of the most massive galaxies at z = 10 in 1 Mpc3 volume in the simulationsby Parry et al. (2014), in preparation.

more than a decade ago (Ricotti et al. 2001,2002a,b). Since then, many progresses havebeen made on understanding the formation ofthe first stars and Population II star formationin molecular clouds (e.g., Wise et al. 2012;Muratov et al. 2013, Parry et al. 2014, in prepa-ration)

Today’s adaptive mesh refinement (AMR)simulations have sufficient resolution to re-solve the multi-phase ISM and molecularclouds in dwarf galaxies. However, the mainqualitative results found in early works havebeen confirmed by modern AMR simulations.Here is a concise summary of the main quali-tative results:

1. Star formation on cosmological volumescales is self-regulated by feedback loopsand is nearly independent of the star for-mation efficiency assumed in molecularclouds. This is illustrated in Figure 1(left).

2. Population III star formation becomesrapidly subdominant with respect to pop-ulation II star formation unless the IMF

of population III stars is such as to leadto direct collapse of black holes with littleor no metal pollution (Ricotti & Ostriker2004b,a)[see Figure 1(right)].

3. Near a mass threshold of 106 − 107 M� asignificant fraction of dark matter halos re-main dark and the scatter of the M/L ratioat these masses is very large.

4. Primordial dwarf galaxies have stars dis-tributed in spheroids (Figure 3) with half-light radii 100 − 500 pc nearly indepen-dently of their luminosity (see Figure 2).

5. Simulations find about 10-100 luminousdwarf galaxies per Mpc3 at z = 10, butis not yet clear whether these simulationshave converged. Population III star for-mation is still implemented using a sub-grid recipe, but the gravitational potentialsat the center of minihalos of mass 105 −106 M� are not resolved with a sufficientnumber of dark matter particles to allow thecollapse of Population III stars that seed thesubsequent Population II star formation.



Data from Kirby et al 2013

Fig. 5. (Left.) Metallicity distribution of stars in six of the most massive galaxies at z = 10 in 1 Mpc3

volume in the simulations by Parry et al. (2014), in preparation. (Right.) Observations of the metallicitydistribution of stars in eight local group dwarf galaxies from Kirby et al. (2011).

2.3. Metals and multi-phase ISM

The improvements in resolution of AMR sim-ulations allow to model the physics of multi-phase ISM and star formation in molecularclouds (see Figure 4). The treatment of starformation and Supernovae (SN) feedback inmolecular clouds, although more realistic thanin previous simulations adopting a Kennicutt-Schmidt law for star formation, is challengingfor several reasons. All the processes importantfor the destruction and formation of molecularclouds need to be included and resolved in thesimulations. Energy injection from SN explo-sions and radiation from massive stars is nec-essary to destroy molecular clouds. Since thescales of molecular clouds are barely resolved(the resolution is about 1 pc) simulations sufferthe “overcooling” problem when the energy ofSN explosions is injected in the dense molec-ular medium. Several approaches have beentaken in the literature to avoid this problem:suppressing cooling, spreading the energy in-jection over neighboring cells, or including theeffect of radiation pressure. In our simulations

we do not include radiation pressure but wealso do not take any other shortcut. The dis-tribution functions of metallicity of stars arein good agreement with observations of localdwarfs (see Figure 5) but current simulationsoverproduce the mean metallicity of the starsfor a given luminosity of the galaxy. The inef-fectiveness of SN feedback in expelling met-als from the ISM may also reduce the impactof the first galaxies on polluting the IGM withmetals (see Figure 6).

2.4. Connection to near field cosmology

The recently discovered population of ultra-faint dwarfs (Zucker et al. 2006; Belokurovet al. 2007; Majewski et al. 2007) in combi-nation with a proper treatment of observationalincompleteness (Koposov et al. 2008) has in-creased the estimated number of Milky Waysatellites to a level that can be more easily rec-onciled with theoretical expectations. For in-stance, the suppression of dwarf galaxy for-mation due to IGM reheating during reioniza-tion (Babul & Rees 1992; Hoeft et al. 2006;


Metals

Gas 300 pc ≈ 0.12 Rvir

10 kpc ≈ 5 Rvir


Fig. 6. Portrait of four of the most massive galaxies (∼ 108 M�) at z = 10 in 1 Mpc3 volume in thesimulations by Parry et al. (2014), in preparation. The top panels show the gas density in a region of 300 pcand the bottom panels the gas metallicity in a region of 10 kpc around the same galaxies with color codingas indicated by the labels.

Okamoto et al. 2008; Ricotti 2009), in conjunc-tion with a strong suppression of star formationin small mass pre-reionization dwarfs, is suffi-cient to explain the observed number of MilkyWay satellites. We can now answer perhaps amore interesting question: what is the mini-mum mass that a galaxy can have? Answeringthis important question requires a good under-standing of the feedback mechanisms that reg-ulate the formation of the first galaxies beforereionization, the details of the process of reion-ization feedback itself and understanding theproperties of the dark matter on small scales(Polisensky & Ricotti 2011). Current observa-tional data supports the thesis that a fraction ofthe new ultra-faint dwarfs recently discoveredin the Local Group are in fact fossils of the firstgalaxies. Although destruction and tidal trans-formation of the Milky Way satellites and theidentification of their masses at formation re-main open questions, the age and metallicitiesof the stellar populations of several ultra-faints

are consistent with their identification as fos-sils (Brown et al. 2012).

Theoretical modeling tells us that the pop-ulation of first galaxies at z = 10 has differ-ent properties from the subset of surviving fos-sils at z = 0. In Bovill & Ricotti (2011a,b) wepresented a method for generating initial con-ditions for LCDM N-body simulations whichprovide the dynamical range necessary to fol-low the evolution and distribution of the fos-sils of the first galaxies on Local Volume scales(5-10 Mpc) We show that he stellar propertiesof most of the ultra-faint dwarfs and classicaldSph are consistent with those expected for thefossils and predict the existence of a yet un-detected population of extremely low surfacebrightness dwarfs. Figure 7, taken from Bovill& Ricotti (2011a), shows a comparison of theproperties of simulated fossils (shaded areas)and dwarfs galaxies in the Local Group. Theasterisks are dIrr, crosses are dE, filled cir-cles and triangles are the classical dSphs in


the Milky Way and M31 respectively, and theopened circles and triangles are the ultra-faintpopulations. We color the observed dwarfswhose half-light radii are are inconsistent withour simulations in green. The magenta con-tours show the undetectable fossils with sur-face brightness below the detection limit (solidblack lines) of the Sloan Digital Sky Survey(SDSS) (Koposov et al. 2008). In both panels,the dashed black lines show the trends fromKormendy & Freeman (2004) for luminous Sc-Im galaxies. A summary of the main propertiesof the fossils of the first galaxies at z = 0 is asfollows:

1. Surviving fossils are anti-biased at z = 10and tend to be underluminous for a givenhalo mass: most classical dwarfs leave inhalos with vmax > 20 km/s, thus reioniza-tion does not have a strong effect on theirstar formation history.

2. The observed population of dwarfs withrh < 100 has properties incompatible withsimulated fossils. Their properties are mostlikely shaped by tides.

3. Models in which some of the ultra-faintdwarfs are fossils of the first galaxies agreewell with observations of the ultra-faintdwarfs but show some tension at the brightend of the satellite luminosity function(classical dwarfs and dIrr). A numerouspopulation of ultra-faint dwarfs also pro-duces an overabundance of bright dwarfsatellites especially in the outer parts of theMilky Way. However, this tension is easedby the expected large spatial extent of theold stellar population that forms a “ghosthalo” difficult to detect and easily strippedby tides around bright satellites.

4. Indeed, the existence of diffuse stellar ha-los around isolated dwarfs is another obser-vational test for the existence of the fossilsof the first galaxies.

3. Summary and conclusions

Near field cosmology appears the most promis-ing avenue to study the epoch of formationof the first stars and galaxies, also in light of

Fig. 7. Surface brightness and half light radius areplotted against V-band luminosity. The shaded con-tours show the distribution for the fossils in a simu-lation by Ricotti & Gnedin (2005) and the overlaidblack symbols show the observed dwarfs. See thetext for the explanation of symbols and lines.

the future surveys that will allow us to probefainter and more distant ultra-faint dwarfs.From the theoretical point of view, the prob-lem seems tractable but much progress needsto be made in our treatment of feedback pro-cesses. The effects of the first black holes andX-ray heating need to be better understood. Inorder to make a connection with detailed obser-vations of stellar abundances in local dwarfs,the treatment of chemical enrichment needs tobe improved by tracing the production of dif-ferent elements, and their transport and mixingin the interstellar and intergalactic medium.

From our studies to connect the fossils ofthe first galaxies to the satellites of the MilkyWay, a good agreement is found between thefossils and the ultra-faints, however a problempersists at the bright end of the satellite lumi-nosity function. The fossil light in the mostmassive satellites appears to exceed the obser-vations with an excess of bright satellites es-pecially in the outer parts of the Milky Way.This problem is alleviated if we hypothesizethat the fossil stars are dynamically hot pro-ducing an extended “ghost stellar halo” with


surface brightness below the detection lim-its or stripped by tides for satellites close tothe Milky Way. It is thus conceivable that themass-to-light ratio in small mass halos in nota monotonic function of the halo mass, andthus some of the classical dwarf satellites of theMilky Way may not reside in the most massivesubhalos. This conjecture would also address apotential problem with the density of the mostmassive Milky Way satellites noted by Boylan-Kolchin et al. (2011). However, Polisensky &Ricotti (2013) showed that this problem is verysensitive to cosmological parameters, in partic-ular to ns and σ8 that determine the power atsmall mass scales and thus the redshift of viri-alization and density of the Milky Way satel-lites. Adopting the most recent cosmologicalparameters and considering the uncertainty onthe mass of the Milky Way, the problem is notsevere, even assuming NFW profiles for thesatellites.

Acknowledgements. MR’s research is sup-ported by NASA grant NNX10AH10G and NSFCMMI1125285. This work made in the ILPLABEX (under reference ANR-10-LABX-63) wassupported by French state funds managed by theANR within the Investissements d’Avenir pro-gramme under reference ANR-11-IDEX-0004-02.

References

Babul A., Rees M. J. 1992, MNRAS, 255, 346Belokurov V., Zucker D. B., Evans N. W., et al.

2007, ApJ, 654, 897Bovill M. S., Ricotti M. 2009, ApJ, 693, 1859Bovill M. S., Ricotti M. 2011a, ApJ, 741, 17Bovill M. S., Ricotti M. 2011b, ApJ, 741, 18Boylan-Kolchin M., Bullock J. S., Kaplinghat

M. 2011, MNRAS, 415, L40Brown T. M., Tumlinson J., Geha M., et al.

2012, ApJ, 753, L21Haiman Z., Abel T., Rees M. J. 2000, ApJ, 534,

11Hoeft M., et al. 2006, MNRAS, 371, 401Kirby E. N., et al. 2011, ApJ, 727, 78Koposov S., Belokurov V., Evans N. W., et al.

2008, ApJ, 686, 279

Kormendy J., Freeman K. C. 2004, in Darkmatter in galaxzies, S. D. Ryder et al. eds.,(ASP, San Francisco), IAU Symp. 220, 377

Machacek M. E., Bryan G. L., Abel T. 2001,ApJ, 548, 509

Machacek M. E., Bryan G. L., Abel T. 2003,MNRAS, 338, 273

Majewski S. R., Beaton R. L., Patterson R. J.,et al. 2007, ApJ, 670, L9

Muratov, A. L., Gnedin, O. Y., Gnedin, N. Y.,& Zemp, M. 2013, ApJ, 773, 19

Okamoto T., Gao L., Theuns T. 2008,MNRAS, 390, 920

O’Shea B. W., Norman M. L. 2008, ApJ, 673,14

Polisensky E., Ricotti M. 2011, Phys. Rev. D,83, 043506

Polisensky E., Ricotti M. 2013,ArXiv:1310.0430

Ricotti M. 2009, MNRAS, 392, L45Ricotti M. 2010, Advances in Astronomy,

2010, 271592Ricotti M., Gnedin N. Y. 2005, ApJ, 629, 259Ricotti M., Gnedin N. Y., Shull J. M. 2000,

ApJ, 534, 41Ricotti M., Gnedin N. Y., Shull J. M. 2001,

ApJ, 560, 580Ricotti M., Gnedin N. Y., Shull J. M. 2002a,

ApJ, 575, 33Ricotti M., Gnedin N. Y., Shull J. M. 2002b,

ApJ, 575, 49Ricotti M., Gnedin N. Y., Shull J. M. 2008,

ApJ, 685, 21Ricotti M., Ostriker J. P. 2004a, MNRAS, 350,

539Ricotti M., Ostriker J. P. 2004b, MNRAS, 352,

547Salvadori S., Ferrara A. 2009, MNRAS, 395,

L6Tegmark M., Silk J., Rees M. J., et al. 1997,

ApJ, 474, 1Wise J. H., Abel T. 2007, ApJ, 671, 1559Wise J. H., et al. 2012, ApJ, 745, 50Zucker D. B., Belokurov V., Evans N. W., et al.

2006, ApJ, 650, L41

Documents

Star formation and multi-phase interstellar medium in the ...sait.oat.ts.astro.it/MmSAI/85/PDF/347.pdf · Star formation and multi-phase interstellar medium in the rst galaxies