Observational Evidence of AGN Feedback

35 Pages·2012·1.11 MB·English

Observational Evidence of AGN Feedback
A.CFabian
InstituteofAstronomy,MadingleyRoadCambridgeCB30HA,UK
2
1
0
2 Abstract

r Radiation, winds and jets from the active nucleus of a massive galaxy can interact with its interstellar
p
medium leading toejectionor heating of the gas. Thiscan terminatestar formation inthegalaxy and stifle
A
accretion onto theblack hole. Such ActiveGalacticNucleus (AGN) feedback can account for the observed
8 proportionalitybetweencentralblackholeandhostgalaxymass.Directobservationalevidencefortheradiative
1 or quasar mode of feedback, whichoccurs whenthe AGNisveryluminous, has beendifficult toobtainbut
is accumulating from a few exceptional objects. Feedback from the kinetic or radio mode, which uses the

] mechanical energyofradio-emittingjetsoftenseenwhentheAGNisoperatingatalower level, iscommon
O
inmassiveellipticalgalaxies. Thismodeiswellobserveddirectlythrough X-rayobservationsofthecentral
C galaxiesofcoolcoreclustersintheformofbubblesinthehotsurroundingmedium.Theenergyflow,whichis
. roughlycontinuous,heatsthehotintraclustergasandreducesradiativecoolingandsubsequentstarformation
h
by an order of magnitude. Feedback appears to maintain a long-lived heating/cooling balance. Powerful,
p
jettedradiooutburstsmayrepresentafurthermodeofenergyfeedbackwhichaffectthecoresofgroupsand
-
o subclusters. NewtelescopesandinstrumentsfromtheradiotoX-raybandswillcomeintooperationoverthe
r nextfewyearsandleadtoarapidexpansioninobservationaldataonallmodesofAGNfeedback.
t
s
a
[ Contents

1
v 1 Introduction 2
4
1 2 TheRadiativeorWindMode 3
1
2.1 Radiationpressureondust . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4
4
2.2 AGNwinds . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
.
4
2.3 Galaxyoutflows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
0
2.4 FromthepeaktothelateevolutionofAGNandquasars . . . . . . . . . . . . . . . . . . . . . . 7
2
1 2.5 Mergersorsecularevolution?. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8
:
v
3 TheKineticMode 8
i
X 3.1 Heating/Coolingbalance–Maintenancemodefeedback . . . . . . . . . . . . . . . . . . . . . 16
r 3.2 Cool,ColdGasandStarFormation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17
a 3.3 Theevolutionofcoolcoresinclusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.4 Themostluminousclusters . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20
3.5 Hotgasingroupsandellipticalgalaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21
3.6 TheKineticluminosityfunction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22
4 BaryonprofilesatdifferentmassscalesandAGNfeedback 22
4.1 PowerfulRadioGalaxies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
4.2 SimilaritieswithGalacticBlackHoleBinaries . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
5 Futurestudies 24
6 Summary 25
7 Acknowledgements 25
1
1 Introduction
Ithasbeenrealisedoverthepastdecadethattheblackholeatthecentreofagalaxybulgeisnomereornament
butmayplayamajorroleindeterminingthefinalstellarmassofthebulge. Theprocessbywhichthisoccursis
knownasAGN(ActiveGalacticNucleus)feedbackandittakesplacethroughaninteractionbetweentheenergy
andradiationgeneratedbyaccretionontothemassiveblackhole(theAGN)andthegasinthehostgalaxy. The
possibilityariseswheretheintensefluxofphotonsandparticlesproducedbytheAGNsweepsthegalaxybulge
cleanofinterstellargas,terminatesstarformation,andthroughlackoffuelforaccretion,terminatestheAGN.
Theratioofthesizeoftheblackholetoitsmassivehostgalaxyistinyandsimilartoacoinincomparison
totheEarth.Thefeedbackprocessmustthereforeoperateoverahundredtothousandmillionfoldrangeofscale
(i.e. 108−109). Thedetailsofthefeedbackarecomplexandtheobservationalevidenceisnotalwaysclear.
Theoverallpicturein termsof energeticsisfairly straightforwardand atleast two majormodeshavebeen
identified,differentiatedbythenatureoftheenergyoutflowneartheblackhole. Thefirstistheradiativemode,
alsoknownasthequasarorwindmode,whichoperates,oroperated,inatypicalbulgewhentheaccretingblack
holewasclosetotheEddingtonlimit. Itismostconcernedwithpushingcoldgasabout.Thesecondmodeisthe
kineticmode,alsoknownastheradiojet,ormaintenancemode. Thistypicallyoperateswhenthegalaxyhasa
hothalo(orisatthecentreofagrouporclusterofgalaxies)andtheaccretingblackholehaspowerfuljets. At
thepresentepochittendstooccuratalowerEddingtonfractionandinmoremassivegalaxiesandinvolveshot
gas.
Afurther,ill-understood,modemaybeassociatedwithgiantradiosources,whichrangeinsizeuptoafew
Mpc. The energies in these sources are prodigiousand approach the binding energy of the gas in groupsand
subclusters.
Itiseasytodemonstratethatthegrowthofthecentralblackholebyaccretioncanhaveaprofoundeffecton
itshostgalaxy. Ifthevelocitydispersionofthegalaxyiss thenthebindingenergyofthegalaxybulge,which
isofmassM ,isE ≈M s 2. ThemassoftheblackholeistypicallyobservedtobeM ≈1.4×10−3M
gal gal gal BH gal
(Kormendy&Gebhardt2001;Merritt&Ferrarese2001;Ha¨ring&Rix2004). Assumingaradiativeefficiency
for the accretion processof 10%, then the energyreleased by the growthof the black hole is givenby E =
BH
0.1M c2. ThereforeE /E ≈1.4×10−4(c/s )2. Most galaxieshave s <400kms−1, so E /E >80.
BH BH gal BH gal
Theenergyproducedbythegrowthoftheblackholethereforeexceedsthebindingenergybyalargefactor. If
evenasmallfractionoftheenergycanbetransferredtothegas,thenanAGNcanhaveaprofoundeffectonthe
evolutionofitshostgalaxy.
Fortunatelyaccretionenergydoesnotsignificantlyaffectthestarsalreadyexistinginthehostgalaxy,orthere
wouldnotbeanygalaxiesasweknowthem. Nevertheless,observationalevidenceisreviewedherethatenergy
andmomentumfromaccretionontothecentralblackholecancouplestronglywiththegasfromwhichnewstars
forms.
AGN feedback is a relatively young topic and a wide range of argumentand opinion has been expressed.
One opinion holds that AGN feedback locks the mass of the black hole to that of its host galaxy bulge, and
determines the ultimate stellar mass of the bulge. Another opinion has it as just one of many processes of
comparableimportancein galaxyevolution;with the blackhole – galaxybulgemass correlationmerelybeing
aresultofrepeatedmergers(e.g. Jahnke&Maccio´ 2011). Thereislittleornoevidenceatthepresenttimefor
AGNfeedbackoperatinginlowmassgalaxieswherestellarfeedbackisimportant,orthatitsignificantlyaffects
galaxydisks,orpseudobulges.
TheclearestobservationalevidenceforAGNfeedbackisfoundinthemostmassivegalaxiesknown,Brightest
ClusterGalaxies(BCGs)incoolcoreclustersofgalaxies.Withoutenergyinputthroughkineticfeedback,many
BCGswouldbeyetmoremassiveandappearasbrilliant,giantstarbursts.
AGNfeedbackfeaturesinmanytheoretical,numericalandsemi-analyticsimulationsofgalaxygrowthand
evolution(e.g. Kauffmann&Haehnelt2000;Granatoetal2004;DiMatteoetal2005;Springel,DiMatteo&
Hernquist2005; Boweretal2006;Crotonetal2006;Ciottietal2010;Hopkinsetal2006,Scannapiecoetal
2011). Theyarenotreviewedhere. The9ordersofmagnitudeinphysicalscalemeansthatallsuchsimulations
includesubgridassumptionsandapproximations.
Thereviewbeginswithabriefoutlineofthephysicsbehindtheradiativemode,thendiscussestheeffectsof
radiationpressureondustygas, followedbyAGNwinds, outflowsandAGN evolution. Theradiativemodeis
themostlikelyAGNfeedbackexplanationfortheblackholemass–stellarvelocitydispersionrelation(M−s ,
seeSection2),sinceitreliesontheaccretionbeingradiativelyefficientandclosetotheEddingtonlimit. Itwas
2
probablymosteffectivebackatz∼2−3whenquasaractivitypeakedandgalaxiesweremostgasrich. Much
of the feedback action involves absorption of the quasar radiation, which obscures the AGN itself, so direct
observationalevidenceispatchyatthemoment.
If feedback empties a massive galaxy of gas it will then refill with at least stellar mass loss if isolated, or
withintraclusterplasmaifinaclusterorgroup. Keepingitempty,oratleastkeepingthegashotsoitdoesnot
cool, appears to be the role of the kinetic mode, which is discussed next. This mode gives the most dramatic
observationalexamplesofAGNfeedbackintermsofbubblesinthecoresofclusters.
Energyinjection from powerfulgiantradio galaxiesis treated last. This may have a drastic impact on the
gasingroupsandsubclusters. Observationalevidenceispoor,however,sincemuchofthepowerinrelativistic
electrons (but not protons) is lost in Compton scattering of the Cosmic Microwave Background, the energy
densityofwhichwasmuchhigherinthepast.
The review finishes by consideringwhetherthe long term behaviourof AGN, and the modesof accretion,
parallels the outbursts of Galactic, stellar mass, binary black holes which tend to be radiatively efficient and
windyathighluminosity,andradiativelyinefficientandjettedatlowluminosity.Futureobservationalprospects
forAGNfeedback,whichareverybright,aretreatedinaconcludingsection.
2 The Radiative or Wind Mode
Silk&Rees(1998,seealsoHaehnelt,Natarajan&Rees1998)pointedoutthataquasarattheEddingtonlimit
canpreventaccretionintoagalaxyatthemaximumpossiblerateprovidedthat
fs 5s
M ∼ T ,
BH 4p G2m c
p
wheres isthe Thomsoncrosssectionforelectronscatteringand f is thefractionof thegalaxymassin gas.
T
Thegalaxyisassumedtobeisothermalwithradiusr,sothatitsmassisM =2s 2r/G.Themaximumcollapse
gal
rate,∼2fs 3/G,isequivalenttothegascontent, fM ,collapsingonafreefalltime,r/s ,requiringapowerof
gal
∼ fs 5/G to balanceit which is limited by the Eddingtonluminosity L =4p GM m c/s . The argument
Edd BH p T
is based on energy which is necessary but may not be sufficient for ejecting matter (the rocket equation, for
example,isbasedonmomentum).
Momentumbalancegivesanexpression(Fabian1999,Fabian,Wilman&Crawford2002,King2003,2005,
Murray,Quataert&Thompson2005)
fs 4s
T
M = ,
BH p G2m
p
whichisaboutc/s timeslargerandinstrikingagreementwiththeobservedblackholemassvsstellarvelocity
disperson(M −s )relation(e.g.Gu¨ltekinetal2009)foraplausiblegasmassfraction f ∼0.1.
BH
Thereare severalways to derive the aboveformula. A simple one is to assume that the radiationpressure
fromthe Eddington-limitedquasarL /chassweptthegas, ofmass M = fM , to theedgeofthe galaxy.
Edd gas gal
Balancingtheoutwardradiationforcewiththeinwardoneduetogravitygives
4p GM m L GM M fGM2 fG 2s 2r 2
BH p Edd gal gas gal
= = = =
s c r2 r2 r2 (cid:18) G (cid:19)
T
i.e.
4p GM m f4s 4
BH p
= ,
s G
T
fromwhichtheresultfollows. Thecancellationoftheradiusintheformulameansitapplieswithinthegalaxy.
TheagreementthatthissimpleformulagiveswiththeobservedM −s relationcanbeinterpretedas(weak)
BH
observationalevidenceforAGNfeedback.
3
2.1 Radiationpressure ondust
TheinteractioncannotrelyonradiationpressureonelectronsasinthestandardEddington-limitformula,since
ifthequasarislocallyatitsEddingtonlimitthenitmustbefarbelowtheEddingtonlimitwhenthemassofthe
galaxyisincluded. QuasarsappeartorespecttheEddingtonlimit, see e.g. Kollmeier(2006)andSteinhardt&
Elvis(2010). (King2003doeshoweverinvokesuper-Eddingtonluminosities.) Theinteractionhastobemuch
stronger, either due to a wind generated close to the quasar which then flows through the galaxy pushing the
gas out, or to dust embedded in the gas, as expected for the interstellar medium of a galaxy (Laor & Draine
1993,Scoville&Norman1995,Murray,Quartaert&Thompson2005).Dustgrainsembeddedinthegaswillbe
partiallychargedintheenergeticenvironmentofaquasar,whichbindsthemtothesurroundingpartially-ionized
gas.L isreducedbyafactorofs /s ,wheres istheequivalentdustcrosssectionperproton,appropriately
Edd d T d
weightedforthedustcontentofthegasandthespectrumofthequasar.
We find that s /s is about 1000 for a gas with a Galactic dust-to-gas ratio exposed to a typical quasar
d T
spectrum(Fabian, Vasudevan& Gandhi2008a), droppingto 500for low Eddingtonratio objects. This means
thataquasaratthestandardEddingtonlimit(forionizedgas)isattheeffectiveEddingtonlimit(fordustygas),
L′ , of a surrounding object 1000 times more massive. Both AGN and galaxy are then at their respective
Edd
Eddingtonlimits. IsthisjustacoincidenceortheunderlyingreasonwhyM /M ∼1000?
gal BH
AGNshowindicationsofaneffectiveEddingtonlimitinthedistributionofabsorptioncolumndensities,N ,
H
asafunctionofEddingtonratio,l =L /L foundinseveralsurveys(Raimundoetal2010). Thereisalack
bol Edd
ofobjectswithcolumndensities,N ,intherange3×1021−3×1022cm−2 andl >0.1. Thisisunlikelytobe
H
anobservationalselectioneffectsincesuchobjectswouldbeX-raybright. Anyobjectfoundinthatzonewould
beofgreatinterestasitcouldtestradiativefeedbackondust. Thegasshouldbeoutflowing.
Interstellargasin anAGN hostevolvessuchthatanywhich straysinto a regionwhereL′ >1 is pushed
Edd
outwarde. Gas which is introduced to a galaxy can remain, fuelling both the black hole and star formation,
providedbothL′ andL remainbelowunity. RepititionofthisprocesscoulddriveM /M →s /s =
Edd Edd BH gal T d
10−3.
IftherepeatedactionofradiationpressureondustisresponsiblefortheM −s relationthenitmustcause
BH
thebulgemasstobes /s timestheblackholemass,
d T
fs 4s
M ∼ d.
gal p G2m
p
Foraconstantmass-to-lightratio,thiscorrespondstotheFaber–Jackson(1976)relation.
SinceM =2s 2r/G,then
gal
s 2 2p Gm
∼ p.
r fs
d
Feedbackshouldshapeboththeblackholeandthegalaxybulgeandmayevenleadtosomeaspects(e.g.s 2(cid:181) r)
of the Fundamental Plane (Faber et al 1987; Djorgovski et al 1987). (It is curious that the above value for
s 2 ∼10−8cms−2 isclosetothefiducialaccelerationa inMONDtheory(Sanders&McGaugh2002).
r o
Galaxiesoccurindarkmatterhaloes,whichdefinetheoutergravitationalpotentialwell. Thetotalmassof
thehalocanbeanorderofmagnitudemorethanthatofthestellarpartofthegalaxy.Silk&Nusser(2010)have
shownthatAGNfeedbackmaynotbeenergeticenoughtoejectallthegasfromthehalo,aswellasthegalaxy,
ifthegasmovesatthe(local)escapevelocity.
Inthespeculativeprocessdescribedabove,wherecyclesofAGNactivitypushthegasoutofthegalaxy,then
thegasmayenduptrappedinthehalo. Itisplausiblehoweverthatthesqueezingofthegasduringtheejection
processtriggersstarformation,leadingtoshellsofstarsoneverlarger(bound)radialorbitsasthegalaxygrows.
Thisinside-outgrowthpatternsuperficiallymatchesobservationsof thedevelopmentofthe radiiofearly-type
galaxiessincez∼2(VanDokkumetal2010)justafterquasaractivityhadpeaked.
2.1.1 Opticaldeptheffectsandanisotropy
Theabovediscussionassumesthattheinfraredradiationproducedbytheabsorptionofquasarradiationbydust
isnotheavilytrapped. Ifitisthenthenetradiationpressureisincreasedproportionaltotheopticaldepth,and
therelationshipsbecomemorecomplicated.
4
Thebulkofquasarradiationoriginatesfromanaccretiondiscandhasabipolarradiationpattern. Thisboth
allowsaccretiontoproceedalongthediscplane,fuelledbymergers,coldflows,orjustsecularevolutionofthe
galaxy,whileatthesametimepushingmatteroutstronglyalongthediscaxis. Gasinthebodyofthegalaxyat
100pcormorewillbemostlysweptupalongthataxis,thegaparoundtheequatorpreventingsignificantlarge-
scaletrappingoftheradiation. Thismeansthatgalaxiesgrowingunderstrongradiationfeedbackasenvisaged
abovecouldappearelongatedalongtheradiationaxis.
2.2 AGNwinds
If the main interaction is due to winds, not to radiation pressure, then the wind needs to have a high column
densityN,highvelocityv,highcoveringfraction f,allatlargeradiusr. Thekineticluminosityofawindis
L f r v 3 N
w
= ,
L 2r c N
Edd g(cid:16) (cid:17) T
wherer isthegravitationalradiusGM/c2 andN =s −1=1.5×1024cm−2.Forhighwindpower,L ∼L ,
g T T w Edd
thenifv∼0.1cthenvaluesofr>103r andN∼N areneeded. Ifthewindispressuredriventhenitmightbe
g T
expectedthatthegasisacceleratedwherevisthelocalescapevelocitysor∼(c/v)2r .
g
To produceM (cid:181) s 4 scaling the thrustof the wind needsto be proportionalto the Eddingtonlimit. This
BH
seemsplausibleifthewindisdustyortheaccelerationisduetoradiationpressureactingonresonancelinesin
thegas.Aproblemwithahighvelocitydust-drivenwindisthatdustisunlikelytosurviveclosetotheblackhole
wheretheescapevelocityishigh. ItisnotclearthatwindstrengthisproportionaltotheEddingtonlimitifthe
windisacceleratedmagneticallyby,say,theBlandford-Payne(1982)mechanism.
The commonestway in which AGN winds are observed is by line absorption of the quasar continuum by
intervening wind material. The X-ray warm absorbers commonly seen in Seyfert galaxies (Reynolds 1997,
Crenshaw,Kraemer&George2003)flowingat∼1000kms−1 are insufficient,bya largefactor(Blustin etal
2005). Faster winds are required, such as those seen in UV observationsof BAL quasars (e.g. Ganguly et al
2007;Weymannetal1991)andinX-rayobservationsofsomeAGN(e.g.Poundsetal2003;Reevesetal2009;
Tombesietal2010,Fig.1),withvelocitiesoftensofthousandsof kms−1. Establishingthatthekineticpower
ofthewindissufficienthasprovendifficult:iftheevidenceofthewindisfromblueshiftedabsorptionlinesthen
obtainingthecoveringfractionandradiusofthewindrequiresindirectarguments.Tombesietal(2012)estimate
thatthemassoutflowrateexceeds5%ofthemassaccretionrateandthatthelowerlimitonthekineticpowerof
theoutflowsinindividualobjectsrangesfrom1042.6−1044.6ergs−1.
Careful work on some quasars (Dunn et al 2010, Moe et al 2010, Saez et al 2010) has established wind
powersat 5–10 per cent of the accretion power, which is sufficientto eject gas from a galaxy. This is backed
up by a range of less direct estimates. An understandingof the overalleffectof powerfulwinds also requires
estimatesoftheirlongevity.
Generally,goodevidenceforAGNwindsoccursinunobscuredAGN,wheretheUVspectrumcanbedirectly
seen. Thereisthenlittle coldgasalongourlineofsightinthehostgalaxytobesweptout,eitherbecausethis
hasalreadyoccuredorthereislittlecoldgasinthefirstplace.Wheresignificantcoldgasispresentinthegalaxy,
the intrinsic AGN spectrummaybe blockedfromview and feedbackis inferredfromthe velocityfield ofany
outflowinthegalaxy.
2.3 Galaxyoutflows
EvidenceofAGNfeedbackisclearlyseeninsomegalacticoutflows. Galacticwindsandstarburstsuperwinds
(Veilleux, Cecil & Bland-Hawthorn 2005; Heckman et al 2000, Strickland & Heckman 2009; Weiner at al
2009) can range from tens to over a thousand Solar masses per year with velocities of a few 100kms−1 for
thecoolcomponents. Mostofthelowervelocitywindsareconsideredtobepoweredbystellarprocessessuch
as supernovae. Identifyingthe effects of AGN feedback in outflows often relies on observing higher velocity
(e.g. >500kms−1) componentsandan outflow powerexceedingthatpredictedbyany centralstarburst. The
detailsarenoteasytodiscern,noristhereyetasimplecleardividinglinebetweenstar-andAGN-drivenoutflows.
500kms−1 is∼1keVperparticleanddifficulttoachievewithstellarprocessesinlargemassesofcoldmolecular
gas. There should of course be a powerful AGN at the centre of the galaxy. High accretion requires a high
5
UFOs
0
3
5
n (%) 20 c 0
o
cti
a
Fr 5
−
0 0.5 1 2 5
1
Observed Energy (keV)
0 0.1 0.2 0.3 0.4
Blue−shift velocity (v/c)
Figure1:Left:DistributionofwindvelocitiesinferredfromfromX-rayabsorptionfeaturesinlowredshiftAGN
(Tombesiet al 2010). Right: Blueshifted X-ray absorptionfeaturesin the most luminouslow redshift quasar,
PDS456atz=(Reevesetal2009).Notethe9keVabsorptionfeatureintherestframeofthequasar,presumedto
beduetoFeXXV(restframe6.7keV)blueshiftedby0.3.Lower:Neutralgasvelocitymapinthequasar/merger
objectMrk231(Rupke&Veilleux2011).
6
fuellingratewhichoftenleadstohighobscurationoftheAGN.Theobscurationbythesurroundinggas,makes
observationsoftheUVandsoftX-raybandswhereabsorptionfeaturesaremostreadilydetected,moredifficult.
AnimportantobjectwhereboththeAGNandoutflowareseenisthelowredshift,z=0.04,quasar/merger
Mrk231. Rupke&Veilleux(2011;Fig.1)mapastrongoutflowinitwithavelocityof∼1100kms−1 andan
outflowrateof420M⊙yr−1,severaltimesgreaterthanthestarformationrate. Theoutflowpowerisaboutone
percentofthebolometricluminosityoftheAGN(seealsoFerruglioetal2010;Fischeretal2010).
Theoptical/UVspectrumofMrk231showsthatit is a lowionizationBAL (LoBal)quasarwith strongad-
ditional absorption (Smith et al 1995). A study of FeLoBAL quasars at 0.8<z<1.8 by Farrah et al (2012)
concludesthatradiativelydrivenoutflowsfromAGNacttocurtailobscuredstarformation(inferredfromtheIR
luminosity)inthehostgalaxiesofreddenedquasarstolessthan∼25%ofthetotalIRluminosity.
Sturmetal(2011)haveusedHerschel-PACStoobservethefar-infraredspectrumoftheOH79m mfeaturein
severallowredshiftUltra-LuminousInfraredGalaxies(ULIRGs). Theyfindhighvelocitiesabove1000kms−1
andmassoutflowratesofupto1200M⊙yr−1 intheAGNdominantones. Thegasdepletiontimesrangefrom
106−108yr. TheirresultleaveslittledoubtthatmassiveoutflowsaregeneratedbyAGN.
TherearemanyrecentreportsofoutflowsfromgalaxieshostingAGN.1000kms−1 outflowshavebeenseen
either side of an obscured quasar at z=0.123 (Greene, Zakamska & Smith 2011), in massive post-starburst
galaxiesatz∼0.6(Tremonti,Moustakas&Diamond-Stanic2007),andcovering4–8kpcofanUltraLuminous
InfraredGalaxy hostingan AGN at z∼2 (Alexanderet al 2010). A regionovera luminousquasar at z=2.4
showsstarformationsuppressed,asinferredfromdecreasedHa emission,wheretheoutflowvelocity,deduced
from[OIII]emission, is highest(Cano-D´ıazetal 2011). Absorptionfeaturesin the spectrumof a background
quasarshiningthroughthehalo108kpcoutfromaz=2.4quasarrevealextremekinematicsinmetal-richcold
gas(Prochaska&Hennawi2009).
Aspectacularexampleisthe1300kms−1 outflowinaredshift6.4quasarrevealedbybroadwingsofthe[CII]
emissionline(Maiolinoetal2012).Thekineticpowerintheoutflowis∼ 2×1045ergs−1 andthe2×1010M⊙
moleculargascontentofthehostgalaxy,inferredbyCOobservation,isejectedin<107yr.
Strong outflows are also seen in radio galaxies at both low (e.g. Morganti et al 2007) and high redshifts
(Nesvadbaetal2008,2011). Theaboveresultsareaforetasteofwhatcanbeexpectedoverthenextfewyears
asinstrumentationandtechniquesimprove.
2.4 From the peak to thelateevolutionofAGNand quasars
AnimportantdiscoveryofthepastdecadewasthecosmicdownsizingofAGN.Themostluminousandmassive
AGN were mostnumerousatredshiftsof2–2.5,the less-luminouspeakedat successivelylowerredshiftswith
the least luminous peaking around redshift one. Downsizing of AGN was first seen in X-ray surveys (Ueda
et al 2003, Hasinger, Miyaji & Schmidt 2005; Barger et al 2005) where the nucleus stands out clearly above
thesurroundinggalaxyineventhelow-luminosityobjects. Laterworkinopticalandotherbandsconfirmsthis
picture.DownsizingisalsoseeinradioAGN(Rigbyetal2011).
Thebehaviouris the oppositeof whatis simplypredictedin a hierarchicalCDM universe,wherethe most
massiveobjects(clustersofgalaxies)formlast. Itindicatesthatsomethingisquenchingquasarbehaviourand
themostwidelyacceptedsolutionisthatitisduetoAGNfeedback.Inmanymodels,massivegalaxiesmergeto
generateamassiveblackholesurroundedbydensegas. Thegasfeedsbothstarformationandanactivenucleus
(e.g. Sandersetal1988). ThepoweroftheAGNblowsthegasawayleavingared,deadellipticalgalaxy(e.g.
Springel,DiMatteo&Hernquist2005).
Studiesofthecoloursofellipticalgalaxiesindicatethatgalaxiesmoveonacolour–magnitudediagramfrom
thebluecloudofstarforminggalaxiestotheredcloudofdeadones.InterestinglymostofthehostsofAGNare
foundin the “greenvalley” betweenthese two extremes(Nandraet al 2007; Schawinskietal 2007). The rate
atwhichthegalaxieshavechangedincolourcanbededucedfrompost-starburstsignaturesintheirspectra,and
appearstobeafew100millionyears,significantlyfasterthanwouldbeexpectedfrompassiveevolution,where
stellarmasslosswouldaccumulateandleadtolatestarformation.
The observational details of this are however uncertain in local galaxy bulges (Wild, Heckman & Charlot
2010),withsignsthattheblackholefuellingmaylagbehindthestarburst(seeHopkins2011foramodel). Bell
etal(2011)studyingmassivegalaxiesfrom0.6<z<2.2findquiescencetocorrelatepoorlywithstellarmass.
A commonfactor of a quiescentgalaxy is that it has a bulge, with presumablya central black hole consistent
withblackholefeedback. A possiblyimportantuncertaintyondistantAGNhostsiswhethertheyaredustyor
7
not (e.g. Brammer et al 2009). Correcting for dust may remove most AGN from the green valley altogether
(Cardamoneetal2010).
2.5 Mergers orsecularevolution?
Manytheoreticalmodelsforquasarevolutionarebasedongalaxy-galaxymergersbeingthetriggerforgasinfall
onto a black hole. (This is a convenient assumption since the merger rate is predictable from the growth of
large-scaledarkmatterstructure.) Althoughmergersmustoccur,theevidenceforthemtriggeringAGNisweak
atmostredshifts. Searchesforpost-mergerdisruptionsignaturesoftengiveanullresultwhenthehostgalaxies
ofAGNarecomparedwithacontrolsampleoffieldgalaxies(e.gCisternasetal2011;Schawinskietal2011).
MergersarebestseenasatriggerfordistantSubMillimetreGalaxies(SMG,Tacconietal2008;Engeletal2010;
Riechersetal2011)andsomelocalSeyferts(Kossetal2010).Secularprocessesmayneverthelessdominategas
inflowinmassivebulgesatz∼2(Genzeletal2009).
Theevidenceis buildingthatbetweenredshiftsof 2 andthe currentepoch,muchof the evolutionofAGN
issecular(Kocevskietal2011,OrbandeXivryetal2011). Furtherevidenceforthisemergesfromastudyof
theprobabilitythatagalaxyhostsanAGN.Airdetal(2011)findthistobeapower-lawinEddingtonrate,and
largelyindependentofmass(seealsoKauffmann&Heckman2009foradiscussionofEddingtonratiosatlow
redshiftandAlexander&Hickox2011forareviewofblackholegrowth).
Secular evolution has implications for the spin of black holes, which are then likely to be high (Berti &
Volonteri2007). Hints thatmost accretion takes place onto spinningblack holes, with consequenthighradia-
tive efficiency, h , have emerged from application of Soltan’s (1982) argument relating the energy density of
quasar/AGNradiationtothelocalmeanmassdensityinmassiveblackholes.
E(1+z)=hr c2,
BH
whereE istheenergydensityinradiationfromaccretion,zisthemeanredshiftatwhichtheenergyisradiated
and r is the mean smoothed-out density in black holes at the present epoch. The equation is independent
BH
ofcosmologicalmodelandreflectsthefactthatboththeblackholemassandenergyradiatedremainandscale
togetherapartfromthe(1+z)redshiftfactorthatmustbeappliedtotheradiation. Applicationofthisformula
to the X-ray backgroundor quasar counts etc, usually yield a value for h of 0.1 or more (Fabian & Iwasawa
1999;Elvis,Risaliti&Zamorani2002;Marconietal2004;Raimundo&Fabian2009). Thisishigherthanthe
efficiencyofanon-spinningblackholeh =0.057andconsistentwithmoderatetohighspin.
Mergersmaystillbethetriggerforthequasarpeakatredshiftsof2–3. Whetheramergeriswetordry(gas
richorpoor)canhaveasignificanteffectonthefinalmergerproduct,ascanhowandwhetherthemassiveblack
holesofthemerginggalaxiesscoursthefinalgalaxycoreornot(Kormendyetal2010).
ThepictureemergingfrommanyobservationsofmassivegalaxiesandAGNisofradiativefeedbackbeing
animportantprocesswhentheAGN/quasarwashighlyluminousandwithinabouttwoordersofmagnitudeof
the Eddingtonlimit. For massive galaxies, this highlightsthe redshiftrangeof the quasar peak. We nowshift
attention to low redshifts and the most massive galaxies at the centres of clusters and groups. They generally
donothostluminousAGNorquasars. Theydohostthemostmassivesupermassiveblackholes,andareoften
activeradiosources. Feedbacktakesplaceherethroughthekineticmodeinvolvingjetsactingonhotgas.
3 The Kinetic Mode
The more massive galaxies at the centres of groups and clusters are often surrounded by gas with a radiative
coolingtimeshortenoughthatacoolingflowshouldbetakingplace(Fabian1994). TheX-raysweseeindicate
alargeradiativelossandmasscoolingratesoftens,hundredsoreventhousandsof M⊙yr−1:
2Lm m
M˙ = ,
5 kT
witha(factorroughlytwo)downwardcorrectionforgravitationalinfallifthehotgasflowsinwardasaconse-
quenceofcooling. m misthemeanmassperparticleofthegasoftemperatureT andListheluminosity(mostly
emittedintheX-rayband).
8
SomerelevantgaspropertiesofasmallsampleofobjectsareshowninFig.2,rangingfromthehighlumi-
nosity cluster A1835throughthe X-raybrightestcluster in the Sky, A426(the Perseuscluster), the low-mass
cluster A262 cluster to the Milky-Way mass elliptical galaxy NGC720. All the clusters show a large central
temperaturedropwithintheinner100kpcandallobjectsshowaradiativecoolingtimedroppingbelow109Gyr
withintheinner10kpc. Anapproximatemasscoolingrate,intheabsenceofanheatsource,canbededucedby
dividingthegasmasswithinachosenradiusbythecoolingtimeatthatradius.Ifacoolingflowisoperating,the
masscoolingratesneedto beworkedoutcumulativelyincludinggravitationalworkdone, whichwillincrease
theratesbyafactorof1.5–2,dependingonthedetailsoftheprofiles.
The mass cooling rates are such that the clusters should be significantlygrowingtheir stellar mass now, if
radiativecoolingisuninhibitedandthecooledgasformsstars.Observationsdorevealsomestarformationtaking
place,andA1835mayhavethehigheststarformationrateinalowredshiftBCG(∼125M⊙yr−1;Egamietal
2006),butitdoesnotequaltheuninhibitedmasscoolingratewhichis∼1000M⊙yr−1. Onlyiftheinitialmass
function(IMF)ofthestarformationprocessinthesesystemsfavouredlowmassstarscouldtherebesufficient
stars. Thehighpressureenvironmentinaclustercore,wherethethermalpressureisabout1000timesthatofthe
interstellarmediumoftheMilkyWayhasbeeninvokedasanexplanationforlowmassstars,duetoitseffecton
theJeansmass(Fabianetal1982).MostobservationsoftheIMFinawidevarietyofobjectssupportauniversal
IMFwhichdoesnothavemostofitsmassinverylowmassstars,whichwouldberequiredhere. VanDokkum
&Conroy(2010)dohoweverfindanIMFrichinlow-massstarsinasmallsampleofnearbyellipticalgalaxies
(seealsoCappellarietal2012),sothecasemaynotyetbecompletelyclosedforlow-massstarsplayingarole.
ThecentresofA1835andthePerseusclusterdocontainextensivereservoirsofdustyatomicandmolecular
gas,themassofwhichcouldbetheendresultofasignificantcoolingflow,exceptthatitwouldnotthenbeclear
wherethedustandmoleculesformed. Itisgenerallyconsideredthatdustcannotformspontaneouslyindiffuse
cooledgas,whichispresumablyrequiredfirstinordertothenformmolecules. TheX-rayrichenvironmentin
clustersdoeshowevermeanthattheH− routemaybeopenformoleculeformationincoldgas. Dustformation
fromcoldmoleculargashasbeenproposedinthissituation(Fabian,Johnstone&Daines1994)butnodetailed
calculationhasbeenattempted. Thesituationmaychangenowthatsignificantamountsofdustandmolecules
havebeenfoundinsomeveryyoungsupernovaremnants,suchasSN1987A(Matsuuraetal2011)andtheCrab
(Loh,Baldwin&Ferlandetal2011).
XMM-NewtonReflectionGratingSpectrometerobservationsprovidedcrucialinformationagainstasimple
coolingflow modelin thattheyfailed toshow thestronglinesexpectedfromFeXVIIas thegascooledbelow
0.7 keV (Peterson et al 2001; Tamura et al 2001). Detailed fits (Peterson et al 2003)indicated that there was
muchlessgasbelowonethirdoftheouterclustergastemperaturethanwouldbeexpectedin asteadycooling
flow. Eithersomethingwasheatingthegas,orthegaswassomehowdisappearing. Aswillbediscussedbelow,
bothoftheseoptionsareprobablyinvolved.
ThelikelyheatsourceistheAGNintheBCGatthecentreofthecoolcore. Almostallhaveanactiveradio
source(Burns1990,Sun2010). HeatingbythecentralAGNwassuggestedearlybyPedlaretal(1990),Baum
& O’Dea (1991), Tabor & Binney (1994) and Tucker & David (1997). Later work shows that the correlation
betweenradiopowerandcoolingluminosity(ameasureoftherateatwhichgascoolswithinafiducialcooling
radiuswheret is7Gyr)ispoor(Voigt&Fabian2004)andthejetsandthusthekineticpowereitherhastobe
cool
highlysporadicorextremelyradiativelyinefficient(seelater).
Thegeneralconsensusnowisthatthemassiveblackholeatthecentreofthegalaxyisfeedingenergyback
intoitssurroundingsataratebalancingthelossofenergythroughcooling(forreviewsseePeterson&Fabian
2006,McNamara&Nulsen2007,Cattaneoetal2009).
SeveralstepsinthisfeedbackprocessareclearlyseeninX-rayandradioobservations. Muchoftheaction
is spatially resolvedand the gasoptically thin. The accretionflow ontothe blackhole generatespowerfuljets
whichinflatebubblesofrelativisticplasmaeithersideofthenucleus.Thebubblesarebuoyantintheintracluster
orintragroupmedium,separatingandrisingasanewbubbleforms(Churazovetal2000;McNamaraetal2000).
A study of the brightest 55 clusters (Fig. 3; Dunn & Fabian 2006, 2008) originally showed that over 70% of
thoseclusterswherethecoolingtimeislessthan3Gyr,thereforeneedingheat,havebubbles;theremaining30%
haveacentralradiosource. Thisimpliesthatthedutycycleofthebubblingisatleast70%. Updatingthatwork
now indicates (Fig. 3) that the bubble fraction is 19/20 objects (the odd one out – the Ophiuchuscluster – is
undergoingamergerintothecore).Whenprojectioneffectsareconsidered,sincebubblesalongthelineofsight
willbedifficulttodistinguish,thecorrespondingbubblingfractionis>95%. Notethatobjectscannotshiftto
theleftinthisdiagramontimescaleslessthant ;theycanshifttotherightonashortertimescalebutthismust
cool
9
)
V 1014
e A1835 10 Perseus 1012 G
k 10 a
( s
e 1013 m
r
u a
rat 5 1012 5 1011 ss
e (
p M
m
e 1011 1010 ☉)
T
103 M
)
r
(y 1011 1010 gas
e /
m 102 ct
o
ti 1010 ol
g (
n M
i
l ☉
oo 109 109 y
C 102 101 r-1
)
10 100 10 100
V) 1012 1012
e A262 1 NGC 720 G
k 1011
a
(
re 2 1010 109 s m
u
a
t
a 109 s
er 106 s
(
mp 1 108 M
☉
Te 107 0.1 103 )
102 M
r) 102
e (y 1010 1011010 101 gas /
m 100 ct
o
ti 100 109 10−1 ol
g (
n 109 10−2 M
li 10−1 ☉
oo 108 10−3 y
C 10−2 10−4 r-1
)
1 10 100 0.1 1 10 100
Radius (kpc)
Figure2:Gastemperature(+),cumulativegasmass(joineddots),M (<r),radiativecoolingtime(x),t (r),
gas cool
andmasscoolingrate(-),M˙ =M (<r)/t (r)),whererisradius,forA1835(Schmidtetal2001;McNamara
gas cool
etal2006),A426(thePerseuscluster;Fabianetal2006),A262(Blantonetal2004;Sandersetal2010a),and
Milky Way mass elliptical galaxy NGC720 (Humphrey et al 2011), composite courtesy of J. Sanders. The
temperaturesshownherearedeprojectedvaluesandassumesinglephasegas.Spatialandfurtherspectralstudies
oftenshowittobemultiphasenearthecentre.
10