Table Of Content

A&A manuscript no. ASTRONOMY (will be inserted by hand later) AND Your thesaurus codes are: ASTROPHYSICS 02(02.13.5; 09.13.2; 12.03.4) Quasars and Galaxy Formation Joseph Silk1 and Martin J.Rees2 1 Institute of Astronomy, Cambridge, UK, Institut d’Astrophysique de Paris, France, and Departments of Astronomy and Physics, University of California, Berkeley, CA 94720 USA 2 Instituteof Astronomy, Cambridge, UK 8 9 Received October 9, 1997 / Accepted 9 1 n Abstract. Theformationofmassiveblackholesmaypre- according to which primordial clouds fragment into stars a cede the epoch that characterises the peak of galaxy for- with an initial mass function that varies between being J mation, as characterized by the star formation history bottom-heavy, top-heavy or even normal (that is, solar 4 in luminous galaxies. Hence protogalactic star formation neighbourhood-like), depending on the observations that maybeprofoundlyaffectedbyquasar-likenucleiandtheir are being interpreted. The quasar distribution tells us di- 1 v associated extensive energetic outflows. We derive a rela- rectly that at least some massive black holes form early. 3 tion between the mass of the central supermassive black Indeed, the quasar comoving density peaks at z >2, and 1 hole and that of the galaxy spheroidal component, and only declines at z > 3 (Shaver et al. 1996); on the other 0 comment on other implications for galaxy formation sce- hand, the peak of galaxy formation occurs at z ≈ 1.5 1 narios. (Madau et al. 1996;Connolly et al. 1997),although there 0 8 is uncertaintyabout the effects of extinctionin leading to 9 Key words: galaxy formation: supermassive black holes anunderestimateofthegalaxyluminosityathighredshift. / – quasars: outflows Infact,theknownquasars,ortheirdeadcounterparts, h p are likely to be within the cores of at least 30 percent of - these galaxies. To see this, note that combining the in- o tegrateddensity in quasar lightwith the assumption that r t 1. Introduction quasarsradiateatorneartheEddingtonlimityieldsanes- s a timateoftypicaldeadquasar(orblackhole)mass(Soltan It is generally assumed that the first objects to form in v: the universe were stars. However this is by no means as- 1982;Chokshi & Turner 1992)as 107−108M⊙. Whether i one actually could observe an AGN component in high X sured. There is general agreement among theorists that, redshift galaxies depends sensitively on the adopted life- if cosmic structures form hierarchically (’bottom up’), as r time of the active phase: higher redshift helps. The ob- a in cold dark matter (CDM) models, the first baryonic cloudshavemassesintherange105−106M⊙,andthatthe served correlation between massive black holes and dynamically hotgalaxiesthen suggeststhat mosthotgalax- characteristic mass subsequently rises. But it is actually ies, amounting to of order a third of all galaxies in terms not obvious that these clouds would undergo fragmenta- ofstellarcontent,couldcontainsucha massiveblackhole tionintostars.Conditionsinprimordialcloudsdifferfrom (Faber et al. 1996). One note of caution would therefore those prevailing in conventional star-forming clouds that be that dynamically hot galaxies probably form system- thefate ofnearbymolecularcloudsisnotareliableguide. atically earlier than most galaxies, and that if starbursts For example, in the absence of magnetic flux, cloud col- characterize their birth, existing high redshift samples of lapsemayhavebeenfarmorecatastrophicthanisthecase such objects may be incomplete. at the present epoch. A massive disk could have rapidly Nevertheless, while the case remains ambiguous, we shed its angular momentum via non-axisymmetric grav- are sufficiently motivated by the possible implications of itational instabilities, and become so dense and opaque a causal connection between quasars and galaxy forma- that it continued to evolve as a single unit. At very high tion to explore in this note the consequences of a cos- redshifts, the inefficiency of atomic and molecular cool- mogonical scenario in which the first objects to form, at ing viaHandH2 excitationsis compensatedby Compton some highly uncertain efficiency, are supermassive black cooling;Comptondragprovidesanadditionalmechanism holes. We discuss how, during subsequent mergers, the fortransferringangularmomentumandallowingcollapse. holes could grow,and exert a feedback on star formation. This outcome seems no less likely, a priori, than the alternativeevolutionarypathwaysfound inthe literature, 2. Model Send offprint requests to: J. Silk 2 Joseph Silk and Martin J.Rees: Quasars and Galaxy Formation 2.1. Early Growth such objects ’quasars’ – noting, however, that the events we are discussing may occur at higher redshifts than the We suppose that the initial black holes form via a coher- typical observed quasars. An explosion model whereby entcollapse.ThisprobablyimpliesMbh ∼> 106βM⊙,,with outflowsfromearlyquasar-likeobjectsledtocooledshells β ∼ 1. Formation of lower mass holes would be less effi- which fragmented into galaxies was originally developed cient, for at least two reasons. Primordial clouds of mass by Ikeuchi (1981; cf. also Ostriker and Cowie 1981); this less than ∼ 109M⊙ are readily disrupted by supernova- ideahas,however,fallenfromfavourastheprincipalmode driven winds (Dekel and Silk 1986). Given the observed forgalaxyformationbecausepost-shockComptoncooling efficiency of black hole formation, the formation of black would lead to excessive spectral distortions of the cosmic holes of mass below ∼ 106M⊙ is likely to be inhibited. microwave background spectrum. We consider here the Moreover if the precursor object forms a supermassive (more localised)effects onthe gas within the protogalaxy star, there would be substantial mass loss. Since typical in which the quasar is embedded. first generation clouds of primordial CDM have masses The effect of a protogalactic wind may be estimated of order 106αM⊙ where α ∼ 1, the total mass going into as follows. We model a protogalaxy as an isothermal blackholeswouldbesignificanteveniftheyformedinonly sphere of colddark matter that contains gasfractionf gas a small proportion of clouds. Of course it is also possi- with density ρ = σ2/2πGr2, constant velocity disper- ble that primordialclouds formsmaller black holes which sion σ, and mass M(< r) = 2rσ2/G. In massive halos, subsequently merge as the hierarchy develops. However T ≈σ2m /3k =4×106K σ/300kms−1 2. A sufficiently this process involves two additional stages of inefficiency p intensewindfromthecentr(cid:0)alquasarcan(cid:1)sweepupthegas (via formation and merging), and we regard it as an im- into a shell, and push it outwards at constant velocity probable pathway. For a typical L∗ galaxy to form from hierarchicalmergingofprimordialcloudsandcontainasu- f L 8π2G 1/3 w Edd v = . permassive black hole, we require that the efficiency f at s (cid:18) f σ2 (cid:19) gas whichthesupermassiveblackholeformed(orequivalently, Inthisexpression,themechanical(i.e.wind)luminosityis theinefficiencyoffragmentationinprimordialclouds)sat- isfies f ∼> 10−5β. takentobeafractionfw oftheEddingtonluminosityL= 4πGcMbhκ−1 = 1.3 × 1046M8ergss−1, Note that fw ≡ How small can f be in order for the consequences to M˙ v2/L = ǫ−1(v /c)2 ≈ 0.01, where ǫ = L /M˙ c2 be of interest? One observes today that many, if not all, out w E w E inf is the radiation efficiency. galaxiescontaincentralsupermassiveblackholes,andthat Expulsionofthisshellrequiresthatitsvelocityshould M = 2×10−3M , where M is the black hole mass bh sph bh exceedthe escapevelocityfromthe protogalaxy:i.e. v > s andM isthemassofthespheroidalcomponent(Magor- sph σ. The condition for this to be the case is that rian et al. 1997; Ford et al. 1997; van der Marel 1997). σ5κ Once supermassiveblackholes areformed, the final black Mbh >αG2c =8×108γ(σ/500kms−1)5M⊙, (1) holemassisenhancedduringthehierarchicalmergingpro- cess, when dynamical friction and dissipative drag on gas where σ500 ≡ σ/500kms−1 and γ−1 ≡ 32π3fw/fgas ∼ 1. can drive supermassive black holes into the center of the If for example γ ∼ 1, black holes could in principle eject developingprotogalaxy.Mergersprovideacontinuingsup- allthematerialfromtheirhostgalaxieswhentheirmasses ply ofgas,andgasdissipationandaccretionfeed the cen- exceed ∼ 107M⊙. If the situation were indeed spherical, tralblackhole.Themostdetailednumericalsimulationsof then, even if the central source switched off, the outflow protogalaxy collapse with cosmological initial conditions would continue to expand into the intergalactic medium that have hitherto been performed demonstrate that an- for up to a Hubble time before stalling due to the ambi- gular momentum transfer is highly effective (cf. Navarro ent pressure.The shellvelocity decreasesvia anexplosive and Steinmetz 1997). Baryons collapse to form a dense, outflow according to massive central clump at the resolution limit of the sim- vs =330(E62/Ω0.05)1/5(1+z)3/10kms−1, ulations, rather than a centrifugally-supported disc on a for an explosion energy, equal to the kinetic energy at galactic scale.This line ofthought at leastsupplies a mo- tivation for exploring (and constraining) the hypothesis breakout, of 1062E62 ergs and an intergalactic medium density equal to 5 percent of the Einstein de Sitter den- that black hole formation and growth, rather than star formation, characterizes the earliest stages of galaxy for- sity (with H0 = 60kms−1Mpc−1), as compared to the binding energy of the gas in a massive protogalaxy of mation.We willarguethatthe valueoff self-regulatesso ∼1060−1061erg.Theambientpressureisnothighenough as to approximately satisfy the observed correlation. to halt the shell, initially moving at breakout at a velocity of ∼ 500kms−1, until z ∼< 1. The shell radius is 2.2. Quasar Winds Rs =4.6(E62/Ω0.05)1/5(1+z)−6/5Mpc, and the fragment mass is (Ostriker and Cowie 1981) Massive black holes, whenever fuelled at a sufficient rate, would display quasar-like activity. For brevity we term Mf =7×1010a410E6−21/5Ω0−.41/5(1+z)−9/5M⊙, Joseph Silk and Martin J.Rees: Quasars and Galaxy Formation 3 where 10a10kms−1 is the sound velocity within the shell. We therefore hypothesize that, in the merger pro- How effective this expulsion would actually be, de- cess leading to the formation of typical galaxies, the hole pendsonthegeometryoftheoutflowandonthedegreeof mass stabilises near the critical value. Hierarchical merg- inhomogeneity of the protogalactic gas. Realistically, the ing, augmented by continuing black hole growth, helps outflowmaybedirectional(probablybipolar)ratherthan maintainthe blackhole-to-spheroidmassratioattheself- spherically-symmetric.Somegasmaysurviveandevenbe regulation level. We suggested in section 1, however,that overpressuredby double radio lobes to form stars (Begel- in the firstbound systems gasmay accumulateinto a sin- man and Cioffi 1989) gle compact unit and evolve into a supermassive hole. If single black holes are indeed favouredoverstar formation It is even possible (Natarajan et al 1997) that swept- in this way, the holes in these first systems would be far up material, after cooling down into a dense shell, may above the ’critical’ value appropriate to such small halos fragment into a class of dwarf galaxies;these would differ withlowvelocitydispersions:inotherwordsM ≫Mcr. fromnormaldwarfs in notbeing embeddedin dark halos, bh bh Star formation would then be inhibited (except in a disc and consequently be more liable to disruption by massive close to the central quasar) until, via mergers, the halos starformation.Henceahigh-massblackholehastwocon- had become large enough to bring the mass of the actual trasting effects. It inhibits star formation in its host halo central hole below the critical mass (which, as we have by blowing gas out; on the other hand, the ejected gas seen, grows faster than linearly with halo mass). There- may eventually pile up in a cool shell that breaks up into after, the scaling would be maintained. small galaxies. If. after mergershad formedhigh-mass halos,the central hole were still above the critical mass, the implica- 2.3. Protogalaxy Core tions would be ominous for galaxy formation. Such a system would end up as a supermassive black hole embed- The implications of (1) are that a massive hole, if it con- dedinalowsurfacebrightnessgalaxy.Aroundthecentral tinues to emit at close to the Eddington rate, could expel host quasar,one would expect to find a cavity of hot gas. gas completely from its host galaxy. (The amount of gas Compton cooling athigh redshift will tend to quench any thathastobeaccretedisinitselfnegligibleinthiscontext: associated extended radio emission. If large radio sources each unit mass of gas accreted into the hole can release were responsible for the hot gas bubbles, then the geom- enoughenergytoexpeloforder(c/v)2 timesitsownmass etry is not a sensitive issue. Indeed, double lobes are as fromthefarshallowerpotentialwellofthehalo.)However, effectiveassphericaloutflowsandaremorereminiscentof the fuelling demands a continuing accumulation of gas in the geometry of the associated hot gas bubbles with the the centre (probably supplied by hierarchical merging). similar energetics of reported Sunyaev-Zeldovich decre- Wecanthereforeinterpret(1)assettinganupperlimitto mentsthathavenoapparentassociatedgalaxyclusterbut the mass of a hole that can exist in a galaxy where star with possibly associated quasars (Jones et al. 1997; Par- formation can proceed efficiently. tridge et al. 1997). Natarajan et al. 1997) argue that one Expression (1) gives a relation between a black hole mightexpecttodetectashellofnewlyformedMagellanic and its surrounding halo. If a significant fraction of the irregular-type galaxies at the periphery of such bubbles. Eddingtonluminosity emergedin ’mechanical’ form,this gives the criterion that the energy liberated on the dy- 3. IMPLICATIONS FOR GALAXY namicaltimescale ofthe halo is equalto the gravitational FORMATION AND SUPPRESSION binding energy. The same expression can be derived by a different argument. The maximum rate at which gas can Near the hole, the ionizing radiation certainly suppresses be fed towards the centre of a galaxy (as the outcome of star formation in normal molecular clouds, which, if of mergers,etc)is∼σ3/G,whereσisthevelocitydispersion density n, survive only at a distance greater than in the mergedprotogalacticsystem. A quasarcould expel ∼ 10(L46/n)1/2kpc from the central quasar. The ultravi- allthisgasfromthegalacticpotentialwellonadynamical olet flux is effective at destroying H2 molecules produced timescale if its mass exceeded a critical value obtained by via H− formation to much larger distances, e.g. the H− requiring that σ5/G=4πGcMbchr/κ. photodissociation rate is ∼ 10−10L46r1−2 s−1 at r1Mpc Thisisindistinguishablefromtheobservedrelationbe- from the quasar whereas the rate of the compet ing H2 tween the masses of central holes and those of their host formation process is ∼ 10−9n s−1 (Tegmark et al. 1997). spheroids (e.g. Magorrianet al. 1997).The observedrela- This would tend to suppress formation of dwarf galaxies tionhasconsiderablescatterbuthints atadependence of out to ∼ 0.3n−1/2 Mpc, If the radiation extends to the black hole on spheroid mass that rises more rapidly than x-ray band, there is however a narrow regime where the linearly:e.g. the MilkyWay hasa centralblackholemass hard ionizing flux, by maintaining a fraction of free elec- of 2.5×106M⊙, whereas M87, with a spheroid mass that trons deep inside the cloud, may actually enhance star is only ∼ 100 larger than that of the Milky Way, has a formation by stimulating H− formation (Haiman, Rees central black hole of mass 4×109M⊙. and Loeb 1996). 4 Joseph Silk and Martin J.Rees: Quasars and Galaxy Formation There are several further noteworthy consequences of erarchybutonlytowardstheendwouldthegasbemostly the hypothesis that supermassive black holes form within exhausted.Perhapsthis wouldhelpexplainthe transition the first subgalactic structures that virialise at high red- with increasing luminosity from coreless spheroids to hot shift, and are in place before most galactic stars have galaxies with cores. formed. AGN activity generates outflows that can inter- act dynamically with the surrounding protogalactic gas We acknowlege helpful discussions with M. Haehnelt as well as provide a possible early flux of hard ionizing and P. Natarajan.The researchof JS has been supported photons.Starformationinthe accretiondisksurrounding in part by grants from NASA and NSF, and that of MJR thebroademissionlineregionoftheAGNislikelytopro- by the RoyalSociety. JS acknowledgeswith gratitude the ceed under conditions very different from those encoun- hospitalityofthe Institut d’Astrophysiquede Pariswhere tered even in starbursts. Star-forming clouds at distance he is a Blaise-PascalVisiting Professor,and the Institute r pc from the supermassive black hole must be dense ofAstronomyatCambridge,whereheisaSacklerVisiting pc enough to avoid tidal disruption, or n∼> 109M8rp−c3cm−3. Astronomer. One infers that the Jeans mass (∝a3/ρ1/2) is reduced to s stellar scale, and also that the specific angular momen- References tumofsuchaclumpisreduced(as∝a2/ρ1/2),relativeto s values in conventional star-forming clouds. Begelman, M. C. and Cioffi, D.F. 1989, ApJ, 345, L21 Chokshi, A. and Turner, E. L. 1992, MNRAS,259, 421 Supposethatthecircum-quasaraccretiondiskisdense Connolly,A.J.,Szalay,A.S.,Dickinson,M.,SubbaRao,M.U. enough to be self-shielding and to be predominantly and Brunner, R.J. 1997, preprint astro-ph/9706255 molecular. If enriched to near-solar abundance level, CO Faber, S. M. et al. 1996, preprint astro-ph/9610055 moleculesandotherspeciesprovideimportantcoolingand Ford,H.C.,Tsvetanov,Z.I.,Ferrarese,L.andJaffe,W.1997, should result in a temperature of order 100K. One then preprintastro-ph/9711299,inIAUSymp.184,TheCentral infers that two of the classical barriers to star formation Regions of theGalaxy and Galaxies, Kyoto, 1997 are likely to be overcome.Hence any gas which falls close Haiman, Z., Rees, M.J. and Loeb, A. 1996, ApJ 467, 522-529 enoughtothecentretobepartofanaccretiondiscshould Jones, M. E. et al. 1997, ApJL, 479, 1-4 convertefficientlyintostars.Outflowfromthesestarsthat Ikeuchi,S. 1981, PASJ, 33, 211-222 formunderthegravitationalinfluenceoftheholeprovides Madau P., et al. 1996, MNRAS,283, 1388 a source of metal enrichment, and may thereby stimulate Magorrian, J. et al. 1997, preprint astro-ph/9708072 van der Marel, R. 1997, preprint astro-ph/9712076, in Proc. more widespread star formation in a protogalaxy. Mag- IAUSymposium 186, Kyoto, August 1997, D. B. Sanders, netic flux will be ejected, the accretiondisk providingthe J. Barnes, eds., Kluwer Academic Publ. conditions for an efficient dynamo. The subsequent tur- Natarajan, P., Sigurdsson, S. and Silk, J. 1997, preprint bulent mixing by supernova remnants and stretching by Navarro, J. and Steinmetz, M. 1997, MNRAS,478, 13 differentialrotationcouldprovideapossibleoriginforthe Ostriker, J. P. and Cowie, L. L. 1981, ApJ, 243, 127 interstellar magnetic field. Partridge,R.B.,Richards,E.A.,Fomalont,E.B.,Kellerman, There are also interesting consequences for the inter- K.I. and Windhorst, R. A.1997, ApJ,483, 38 galactic medium (IGM), since quasar-driven winds could Shaver,P.A.,Wall,J.V.,Kellerman,K.I.,Jackson,C.A.and readily eject enriched material from the shallow potential Hawkins, M. R. S.1996, Nature,384, 439 wells that characterize the earlier stages of hierarchical Soltan, A. 1982, MNRAS,200, 115 clustering. This could account for the apparent presence Tegmark, M. et al. 1997, ApJ,474, 1 of up to 1 percent of the solar abundance of heavy ele- ments even at high z. If this is the case, a time delay is likely between the epoch of most quasar activity and the epoch of the bulk of star formation in the universe. One observesa time delay correspondingto δz ≈1 : studies of quasarshaveconfirmedapeak inthe quasarnumber density between z = 2 and 3, and the star formation history of the universe is found to peak between z =1 and 2. The cores of present-day galaxies may carry traces of the black hole merging history. Megers and accretion will allowthe black hole to growto the criticalvalue as deter- mined by equation (1). Our key prediction is the relation between central black hole mass and spheroid mass. Luminous elliptical cores are best explained by heat- ing associated with binary black hole decay following a majormerger(Magorrianet al. 1997).The presentmodel envisagesthatthisprocessoperatesatallstagesofthehi-