Download 13.1 Galaxy Evolution: Introduction

13.1GalaxyEvolution:Introduction
[slide1]Wenowturnourattentiontothesubjectofgalaxyevolution,whichisatthecoreof
mostofmoderncosmologyandextragalacticastronomy.
[slide 2] First of all, galaxies must evolve for two reasons. They're made out of stellar
populationsandstarsevolve.Also,galaxiesmerge,assemblingeverlargerobjectsthrougha
hierarchical structure formation. So, therefore, there are two aspects to galaxy formation.
One is the assembly of the mass, and the other one is conversion of gas into stars and
energy.Theformerisrelativelywell-understood-thisissomethingwecanmodelverywell
innumericalsimulations-andit'sdrivenbythedarkmatter.Thelatterismuchmoredifficult
becauseitinvolvesmessydissipativeprocesses,includingstarformationitself,thefeedback
of stellar populations, supernovae active nuclei, and so on. And, there we can model
processesuptosomelimit,usingmodernhydrosimulationsbecausetheanalytictheoriesof
starformationdonotexist.
So, even though dark matter dominates the process of galaxy assembly because it's the
dominantmasscomponent,that'snotwhatwesee.Whatweseeisthelight.And,fromthat
wehavetoinferabouttheassemblyofthemassaswell.
[slide3]First,let'stakealookattherelevanttimescalesforgalaxyevolution.Onetimescale
willbetypicalstarburstscalewhichwenowthinkisoftheorderof10to100millionyears.
That also happens to be the timescale of the life of the very massive stars, which are
energeticallythedominantonesandwhichalsoprovidemostofthephotonsthationizethe
interstellarmedium.
Italsoturnsoutthatthisistheestimatedtimescaleofactivegalacticnucleusepisodes,and
it's not obvious whether this is coincidence or not. The internal timescales within galaxies
are commensurate with their free-fall timescales, which would be a few hundred million
years.Andthetimescalesneededformergingandassemblyaretypicallyoftheorderofa
billionyears.And,finally,thereisHubbletimeoftheorderof10billionyears,throughwhich
galaxiesevolve.
Therearedifferentobservationalapproachestothisproblem.Firstofall,wecanlookatour
own Milky Way and try to deduce from its constituents and properties how it may have
formed.Next,wecanlookatnearbygalaxiesandstudythesystematicsoftheirproperties,
as we discussed earlier through Hubble sequence, through use of scaling relations, and so
on. And, finally, we can observe it directly by looking deep, since light travels at a finite
speed;thefurtheroutwelook,thedeeperinthepastwelook.Andso,wecanseegalaxy
evolutionunfoldingonourpastlightcone.
[slide4]Todothis,weneedagoodcombinationoftheoreticaltoolsandobservationaltools.
On theory side, we can use numerical simulations of structure formation to tell us about
massassembly.Wecanalsopredictbehaviorofthestellarpopulationsbecauseweknowa
lotaboutstellarevolution,andsomakingsomeassumptionsaboutinitialmassfunctionof
stars and the actual history of star-formation rate in galaxies, we can predict what their
spectrawouldlooklikeasacompositeoftheseevolvingstellarpopulations.And,thereare
alsohybridsemi-analyticalschemesthatcombinethese.
[slide5]Ontheobservationalside,thereagainarethreeapproaches.First,wecantakedeep
images, the deeper the better, if you want to look in the past over a full range of
wavelengths.And,fromthosewecaninferalotaboutstar-forminghistoriesorgalaxiesthat
maybemergingaswell.Moreimportantapproachisthroughspectroscopy,becausethisis
whatrevealsnotjustredshiftsofgalaxies,butphysicalpropertiesthathappeninthem,such
asstar-formationrates,presenceofanactivenucleus,ifthereisone,etc,etc.And,finally,
wecanlookatthediffusebackgrounds,thecollectiveemissionofallgalaxiesever.Now,that
has a drawback of not knowing which source is where until you actually resolve the
background.But,itbypassesselectioneffects,becauseyougeteverything,whetherornot
individualgalaxiesaredetectableornot.
[slide 6] A very important thing is to be aware of the selection effects. And, they come in
multipleguises.Themostobviousoneisthefluxlimit.Anygivenastronomicalobservation
runsintothesignal-to-noiseproblematsomefaintlevel,andthat'showdeepyoucango.
But,moresubtly,thereisasurfacebrightnessselectionlimitthatforextendedobjects,it's
thenumberofpixelsaboveacertainthreshold.Thatis,notjusthowmuchlightthereis,but
howdiffuseit'sdistributed.Themorecompactgalaxieswouldbemucheasiertodetectthan
thediffuseones.Andso,weknowthat,forsure,thedeeperwelook,wearegettinganever
moreluminousorhighersurfacebrightnessobjects,andwe'remissingtheintrisicallyfaint
onesortheintrinsicallydiffuseones.Ifwedon'tknowwhatwearemissing,thenitwillbe
veryhardtocorrect.But,wecanmakereasonableextrapolationsfromnearerpartsofthe
universe,andseehowthatworks.
[slide 7] Let's refresh our memory of the dynamical evolution – galaxy merging. We've
discussedthisearlier,andseenhowhierarchalassemblyofgalaxiesisaprocessthatkeeps
goingonfromtheearliestdaysoftheuniverse,andweseeiteventodayinmajormergersof
galaxies.Thisobviouslyarrangesthemass,butthankstothedissipationandinfallofgasto
thecenters,itcanalsotriggeraburstofstarformationaswellasfeedingofactivenuclei.
Thus, stellar population evolution of galaxies and their dyamical evolution can be very
deeplyinterconnected.
[slide8]Animportantphysicalprocesshereisdynamicalfriction,anditworksasfollows.If
youhaveamassivebody–inthiscase,say,agalaxy–movingthroughaseaofmasspoints–
stars of another galaxy, say – it will accelerate them. And, as it keeps moving, you'll be
encounteringmoreandmorestarstowhichitwillexertitsacceleration.Thosestarsacquire
acertainvelocity,andthatkineticenergyhastocomeattheexpenseoftheperturber.So,
thestellarpopulation,orpopulationoftestparticlesinwhichthisperturberplunges,gains
kinetic energy, usually in the form of random motions, whereas the perturber loses its
orbitalkineticenergy.Andthiswhygalaxiesthatareinitiallyinparabolicorbitscanmerge.
Incidentally, a good local example is the Magellanic Clouds. They are being torn apart by
tidalinteractionwiththeMilkyWay,andinafewbillionyearsthey'llmergewiththeMilky
Way.Thishasalreadyhappenedtonumerousdwarfgalaxiesinthepast,andweseefossil
evidenceofthatinMilkyWay'shalo.
[slide9]Thetimescalefordynamicalfrictioncanbeexpressedasfollows.Youstartwiththe
velocity – relative velocity of the perturber – divided by the rate in which that velocity is
beingdiminished–decelerationduetotheenergyloss.Aftersometheoreticalcomputation,
you obtain a formula like this. It has several interesting features. The denser the medium,
whichhasbeenperturbed,andthehigherthemassoftheperturber,theshortertimescale,
which is intuitively clear. But also, the timescale gets to be longer if the velocity of the
perturberishigh.And,thereasonforthisisthatperturbersthatzipbyveryfastsimplydon't
have enough time to deposit enough kinetic energy in the perturbee galaxy, so it'll take
manydifferentpassesforthedynamicalfrictiontoactuallydoitsthing.Thisiswhythetime
goesup.
[slide10]We'veseennumericalsimulationsofmerginggalaxiesearlier,andthisisasetof
snapshots from one of those showing what dark matter particles do. In these simulations,
always,eventually,themergerproductlookssomethinglikeanellipticalgalaxy.
[slide 11] But, if you follow the evolution of the gas, you find out that the gas is reacting
muchmoretotheencounter.Andit'slosingkineticenergyveryquicklyandveryeffectively,
meaningithastosinktothebottomofthepotentialwell,whichalsomeansitwillachieve
highdensities.Thisisexactlythekindofconditionsthatcanleadtoburstofstarformation,
ortheycanalsofeedagiantblackholeifthereisonethere.
[slide11]Next,wewilltalkaboutthemodelsofthestellarpopulationevolution.
13.2StellarPopulationSynthesis
[slide 1] Evolution of stellar populations in galaxies is one of the key ingredients of our
understandingofgalaxyevolutioningeneral.
[slide 2] If we look at colors of galaxies nearby, we can immediately see that there is a
bimodaldistribution.Thereisafairlynarrowredpeak,whichisduetotheellipticalgalaxies
and bulges, and there is a fairly broad blue distribution that corresponds to the disks of
spirals.Inmoremodernrenderings,youcanplotcolorsasafunctionofintrinsicluminosity,
and you still see those same two blobs, but they're a little tilted. In the case of elliptical
galaxies,thatridgelineisreallymass-metalicityrelation.Themoreluminousellipticalsretain
more of their chemical evolution products. They're more metal-rich. Metals absorb more
lightintheultraviolet–galaxylooksredder.So,thisislargelyamass-metallicitysequence.
For spirals, there isn't a very simple relation like that. Their colors are a mixture of stellar
ages,starformationrates,aswellasextinctionbydust.
[slide 3] So, we can produce predicted theoretical spectra of evolving galaxies in the
followingfashion.Wehavetomakeanassumptionofwhattheirstar-formationhistoryis.
Thisispurelypre-parameteringthemodel,butwecanmakereasonableguesses.Weneed
to know what is the initial mass function - the distribution of stars by mass when they're
formed-becausestarsofdifferentmassescanevolveatverydifferentpace.Then,foreach
stellarmassandandageweneedaspectrum.So,weneedlibrariesofstellarspectrathat
can be associated with all components of the stellar population at any given age. This is
actuallynotaneasythingtodo,becausewecanobservealotnearusbut,forexample,we
have no spectra of very metal-poor very massive stars because those have burnt out long
timeago.And,againfromtheorytestedbyobservationsofstarclustersandsoon,weneed
stellarpopulationevolutiontracks:howdoesdistributionofstarsandcolormagnitudespace
–theHRDiagram–changesasafunctionoftimeforgivenamountofchemicalenrichment,
andsoon.
Youtakeaquantityofgasandturnitallintostarsinstantaneously–adeltafunctionstarformation rate – distributed according to your initial mass function, and then follow its
evolutionintime.Thisiswhat'sknownasasimplestellarpopulation.Thereisnosuchthing
inreality,althoughglobularclusterscomeclose.Andthen,youcanrepresentnettotalstarformationhistoryofagalaxyasacollectionofthese,andfollowthemap.Onepopularway
ofexpressingstar-formationhistoriesisasanexponentiallydecliningrate,whichisprobably
not a bad overall assumption, averaged over many galaxies. In this case, for early-type
galaxies,thereisalotofstarformationearlyon-theexponentialverysteep.Forthisgalaxy,
it'sveryshallow.Forirregulargalaxiesmaybeessentiallyflatorevenrising.
[slide4]Stellarevolutionisoneofthebestunderstoodsegmentsofastrophysics.Wereally
do know how stars work and evolve, and that's been confirmed again and again through
many decades of observations. There are certainly details that still need to be ironed out,
but,atthelevelthatwecareabouthere,wereallydounderstandthat.Animportantthing
torememberisthatmoremassivestarsevolvemuchfaster.Theyaremoreluminous.They
arealsohotter,sothey'reenergeticallymoreimportant,butnotforlong.
Thus, spectrum of an evolving galaxy will change more rapidly in the blue parts of the
spectrum,thankstotheshortlifetimesofthesemassivestars,andwillbechangingrelatively
slowlyintheredderpartsofthespectrum,wherethemostofthelightmaybecomingfrom,
well,allslowlyevolvingpopulationofredgiants.There'remanystellarpopulationsynthesis
methods out there, and they sometimes disagree on some details, but, by and large, the
agreement's pretty good. One popular set is called GISSEL, which is a library of the galaxy
evolutionspectrabyBruzualandCharlot.Weknowhowstarsevolve.Weevenhavespectra
of lots of different kinds of stars. And, we have some idea of the initial mass function, at
least we measure it locally in the Milky Way, and then we have to make assumptions
whether it's the same in all galaxies at all times. We can test some of those assumptions.
But, then we have to assume star-formation rate, which again is a completely free
parameterinthisexercise.
[slide 5] This is what stellar evolution tracks look like. They're computed from stellar
evolution models. And, as a star evolves, it moves in the color magnitude space. How and
where–dependsverymuch,oralmostentirely,onitsmass,withminordependenceonits
metallicity.
[slide 6] So, we get these stellar evolutionary tracks from theory - our understanding of
stellarstructureandevolutionthat'sprettysolid.Then,weneedtohavetheirspectra.And,
forkindsofstarsthatwecanobservenearus,that'saneasythingtoacquire.Forkindsof
starsthatarenolongerwithus–say,veryyoung,metal-poor,veryhigh-massstars–this
willrequiresometheoreticalmodelingandextrapolation.
Wemakeanassumptionaboutinitialmassfunction,andagain,weunderstandthatinthe
localMilkyWayconditions,butitwouldbeverydifferentintheearlyuniversewhere,say,
starsweremadeoutofhydrogen,heliumandhardlyanythingelse.Infact,webelievethat
wasthecase,thattheinitialmassfunctionofprimordialstarswasverydifferent.And,then
again,wehavetoassumestar-formationhistory.Inreality,weassumesometypesofstarformationhistories,computetheconsequences,compareittotheobservations,anditerate
until we have a model that seems to fit observations, and that's telling us what the likely
star-formationhistoryofthesegalaxieswas.
[slide7]So,starsevolve,theblueoneskeepdisappearingfaster.Galaxieswillbechanging
theircolors,andhereincolor-colorspaceyoucanfollowthetheoreticalbehaviorofevolving
stellar populations. You can also see where the galaxies are. And, indeed, they form
somethingofasequencethatcorrespondstotheyounger,hotterstarsonlateHubbletypes,
older,redderstarsfortheearlyHubbletypes.
[slide 8] Here is a comparison of predictions of three different types of stellar populations
andsynthesismodelsbydifferentgroups.And,itshowspredictedcolorsandmass-to-light
ratios for model galaxies. By and large, they agree very well. The minor disagreements
usually are at very young ages, and there is still some debate actually what is the correct
thingtoassume.But,qualitativelyatleast,weunderstandverywellhowthingsareworking.
[slide 9] And so, this is what evolving spectra of stellar populations look like. This is for a
simple stellar population. Remember, this is a scoop of stars made all at once and let go.
Note, these are logarithmic plots of log flux versus log wavelength, so they really hide the
strongcontrast.And,ifyoustareatcurves,whicharelabeledbytheirage–thosenearthe
toptendtobeyoungestbecauseyoungerstarsaremoreluminous–youfindoutthatinthe
ultraviolet,bluepartofthespectrum,thefluxwillplummetveryquickly,whereasitwould
be changing in the red part of the spectrum, but much slower, which is exactly what we
expect.
[slide 10] Here is a comparison of predicted model spectra of different ages, assuming
differentlibrariesofstellarpopulationevolutiontracksandspectra.And,again,youseethat
different models seem to be in an excellent mutual agreement, which is giving us some
confidencethatweactuallydounderstandhowthisworks.
[slide 11] Another kind of models that they are now more popular are so-called semianalytical or hybrid models, where one can use stellar population synthesis models,
associate them with galaxies that they're being assembled through hierarchical structure
formation either from a numerical simulation or from some statistical description thereof,
andmakepredictionsofhowwilltheircolorschange,andsoon.Thereare,unfortunately,
way too many two-number parameters in these models. And so, they have some modest
success,buttheyillustratethecomplexityofalldifferentthingsthathavetobetakeninto
account,andalotofassumptionsthathavetobemade.
[slide12]Nexttime,wewillturntotheactualobservationsofgalaxyevolution.
13.3:ObservingGalaxyEvolution
[slide1]Nowlet'sseewhattheobservationssayaboutgalaxyevolution.
[slide 2] The simplest thing to do is take images of different wavelength that is
observationally much cheaper than taking spectra, which require longer integration times.
And, this can be used to do galaxy counts as a function of their brightness, color etc. But,
that can only go so far, and to really understand what's going on, redshifts really are
necessary. So, it was only in the recent years that we have obtained sufficiently large
samples of galaxies in deep fields to really reach a good observational understanding of
galaxyevolution.
Thereisoneimportantdichotomy.Youcanthinkofstarformationascomingintwoflavors:
thesimple,unobscuredstarformation,whereyouseethelightfromstellarphotospheresas
theyare,orlightabsorbedandre-radiatedbyinterstellardust,whichnowmovesallofthe
energyintofar-infraredorsubmillimeterregime.So,therearetworegimesinwhichwecan
observegalaxyevolutionandstarformationinthem,andeachofthemhasitsowntoolsand
selectioneffectsandlimitations.
[slide3]Youmayrecall,whenwefirstmentionedsourcecountsasapotentialcosmological
test,thattheevolutionaryeffectsreallymessthingsup.And,theyalwaysworkinthesense
of making counts higher than they would be in the absence of the evolution. Even the
galaxies that are evolving in brightness, say, due to the fading of stellar populations that
weremoreluminousinthepast,andthereforetheywouldbeatthehighermagnitudelevel.
But,therewouldbemanymoreofthem,thus,they'dbemovingtotheleftandproducinga
less declining curve. Likewise, if galaxies are assembled from smaller pieces, there were
more smaller pieces in the past than exactly the same observational effect appears. To
disentanglethese,weneedredshiftsurveys.
[slide4]Still,hereatthedeepgalaxycountsfromtheHubbleDeepFields,andthesedayswe
do this down to about 29th magnitude or thereabouts, which is really spectacular, and by
extrapolationovertheentiresky,theremaybeacoupleofhundredbilliongalaxieswithin
theobservableuniverse.
[slide 5] You may also recall that evolution is expected to appear stronger in bluer
wavelengths,andlesssoinredder,andthat's,indeed,exactlywhat'sseenhere.Now,thisis
now infrared galaxy counts, and they show roughly minor discrepancy, compared to the
muchstrongeronesthatareobservableinthebluerpartsofthespectrum.
[slide 6] But, those galaxies evolve as their stellar populations evolve. Their colors evolve
too,generallygoingfrombluertoredder.However,that'salsocomplicatedbytheredshift.
Thewholespectralenergydistributionismovingfrombluerfilterstotheredderfilters.So,
anygiventimeyoucanlookatcolor-colorspaceandseewhatgalaxieswilldointhere.Now,
you can make use of their complex trajectories, and use those to evaluate redshifts from
colors alone. Those are so-called photometric redshifts. You can think of those as a really
low-resolutionspectroscopy.
[slide7]And,theyseemtoworkremarkablywell,typicallyinatleastfourorfivedifferent
filters,andhereareexamplesofsomeofthemeasurements.Thosearedotswitherrorbars
with models of stellar populations drawn through them. This actually looks too good, and
there could be many different models that can fit the same set of photometric
measurements,butthatcanbealsoevaluatedstatistically.
[slide 8] So, here is a typical plot. Usually, there is a really excellent agreement between
spectroscopicredshifts,doneasacontrol,andpredictedphotometricredshifts.Thestateof
the art is that maybe down within a few percent. However, there are always outliers,
galaxiesforwhichgrosserrorismade,andthat'susuallyduetosomethinglikepresenceof
anactivenucleus,orsomeotherpeculiarhappeninglikethat.
[slide9]Aparticularformofphotometricredshiftsreliesonthepresenceofdeepjumpsin
the spectrum of the galaxy. There are a couple of those. There is the limit of the Balmer
seriesofhydrogen,whichthenresultsinastepofmagnitudearound3,600angstromsinthe
rest frame. So, you can use that by measuring flux in filters blueward of the jump and
redward of the jump. An even stronger effect occurs at the limit of the Lyman series, and
thoseareextremelyusefultofindgalaxiesatveryhighredshifts.Moreover,forthereasons
we'll be discussing later, intergalactic medium absorbs light blueward of Lyman-alpha line.
So, it's really the wavelength of the Lyman-alpha line that serves as an interesting jump
point. This has been used to great effect, in particular by Steidel and collaborators, who
discover large numbers of galaxies of high redshifts, and then study their evolution and
properties.
[slide10]Butagain,colorsandmagnitudeshavetheirlimitations,andredshiftsareneeded.
So,thattheadventofeight-,ten-meterclasstelescopeslikeBLTinChile,orKecktelescopes
in Hawaii, in Subaru, and so on, it became possible to actually do this in an effective way.
Also, there was the development of multifibre spectrographs, which you may recall also
revolutionizedtheredshiftsurveysatlowredshift.And,nowadays,thousandsandhundreds
ofthousandsoffaintgalaxyredshiftshavebeenobtained.
[slide 11] A good winning strategy is to obtain really deep images from space where you
don'thavetoworryabouttheeffectsofEarth'satmosphere.ImagestakenfromHubblecan
gomuchdeeperthanthosetakenfromthegroundwithabetterresolution.Andso,thereis
asetofselectedfieldsinthesky,whereverydeepobservationshavebeenobtained.Hubble
DeepFieldwasthefirstone,followedbytheUltra-DeepField,andChandraDeepField,and
eXtremeDeepField.So,thesearethedeepestwindowsintheuniverseweobtainedsofar.
Once you have these images in a number of filters, you can deploy large telescopes to
measure redshifts of as many galaxies as you possible can, and that's still very much an
ongoingenterpriseinobservationalcosmology.
[slide 12] Here is the first one of those, the Hubble Deep Field with its characteristic B2
bomber shape, and the histogram of redshifts obtained with the Keck telescope. So, even
thosewerethedeepestobservations.Untilthen,thebulkofthesegalaxiesisactuallynotat
theveryhighredshifts,aboutredshift½.Theygobeyondredhiftof1,butnotbyalot.
[slide13]Insubsequentwork,pushingeverdeeper,galaxieswerefoundatredshiftsof5or
evenalmost6,butstillthebulkofthese,evenatthelimitofthepresentdayobservations
with eight- and ten-meter class telescopes, is of the order of unity or less. So, we do not
actuallyprobeevolutionofgalaxiesverydeepthroughdirectmeasurements.Smallnumbers
wedosee,butthenonehastobeawareofselectioneffects.Thecompleteunderstandingis
reallyatredshiftslessthan1.
[slide 14] This was done now by numerous groups, both in north and south, and tens of
thousands of galaxy redshifts have been obtained. The results are usually in a really good
mutualagreement.Thisoneisfromapencilbeamsurveyinso-calledGOODSField,whichis
whereHubbleDeepFieldwas,plusotherobservationssurroundingit.
[slide15]Andso,hereareacoupleinterestingdiagrams.Ontheleft,youseetheredshift
histogram.Andit'sveryspiky.It'snotspikybecauseit'snoisy,it'sspikybecauseofthelargescale structure, that the line of sight intersects filaments, or even clusters and voids, and
that'swhatproducestheobserveddistribution.Ontheright,youseeabsolutemagnitudesin
therestframeofgalaxiesplottedversusredshift.And,youseethereisasharpcutoffthat
correspondstomagnitudelimit.Peoplewhodothesesurveysdecidethattheycanonlygo
down to some magnitude level, say, 24 magnitude, and that maps into different absolute
magnitudesatdifferentredshifts.Sothisisabuilt-inbutwell-understoodselectionoffacts.
Nevertheless,onehastobeawareofit.Naively,ifyoulookedatthis,youwouldconclude
that galaxies of higher redshifts are more luminous. No, you simply only see the luminous
oneswithhighredshifts,youarenotseeingthefainterones.
[slide16]So,thiswasdonefordeepfields,andtheresultisasfollows.Individualgalaxies
cannot be really compared, you need to look at the whole population, and the simplest
description of the entire population is the luminosity function – distribution of galaxy
luminosities.So,ifyoulookatthatindifferentredshiftshells,youfindoutthattheobserved
galaxy luminosity function is very similar to the one we've seen near us. It evolves very
slowly,butyoubegintoseeacoupleofinterestingeffects.Byaboutaredshiftof1/2orso,
thefaintendsteepens.Weseemoreevolutioninintrinsicallyfaintergalaxies.And,second
thingisthatasyoupushdeepenough,youstarttoseebrighteningatthebrightend,which
is what you expect from fading of stellar populations. The steepening was a little bit of a
surprise.Itwasthoughtbeforethatgalaxiesresponsiblefortheexcesscountsintheskies,in
theimaging,areevolvinggalaxiesatlargerredshifts.Thebulkofthemturnouttobedwarf
galaxies at modest redshifts. And, the evolution of galaxies depends very much on their
intrinsicluminosity.Thisisknownasthedownsizing.Atfacevalue,it'sexactlyoppositewhat
youexpectfromahierarchicalstructureformation,whereyouslowlybuilduplargegalaxies,
youexpectthelargergalaxiestobeevolvingfastest,butit'sexactlytheopposite.
[slide 17] Moreover, the evolution of the lumosity function depends on the galaxy
morphology. And, if you can split them in morphological type – say, early and late-type
spiralsandellipticals–youfindoutthatthelater-typegalaxies,thosefurthertotherightin
Hubble sequence – star-forming disks and irregulars – have the strongest evolution.
GalaxiesontheearlierparttoHubblesequenceprettymuchdonetheirevolvingbyabouta
redshift of 1. They do evolve since then, but bulk of the change appears in the faint
populationandalsothelateHubbletypes.
[slide18]Itisonlywhenwereachredshiftsoftheorderoftwoandbeyond,thatwestartto
see clear effects of strong evolutional stellar populations at the very bright end. So, the
genericconclusionthesedaysisthatmostoftheHubblesequencewasprettymuchinplace
byaboutaredshiftof1.And,thingshavebeenevolvingatrelativelymodestpacesincethen.
We'llseesomeotherapproachestothisalittlelater.
[slide 19] Next, we will talk more about some of the observed results, as well as the
evolutioninclusters,asopposedtofield,whichiswhatwejusttalkedaboutnow.
13.4GalaxyEvolution:SomeResults;GalaxyEvolutioninClusters
[slide1]So,wehavestartedtalkingabouttheresultsongalaxyevolutionstudiesingeneral
field.First,wefoundouthowtheluminosityfunctionseemstobeevolvingslightly.Butnow,
let'sseesomeoftheotherresultsthathavebeenobtained.
[slide2]Asimplethingtodoistolookatcolorsofgalaxiesasafunctionofredshift.And,if
we then predict what galaxies of different Hubble types would look like - if there was no
evolution, just a K-correction, which you may recall from way back in Hubble diagrams -
what would measurements look like, given the redshifted spectrum? This is what it's like.
The points are the actual measurements and the three lines labeled with different Hubble
types correspond to the colors that those particular Hubble types would have at a given
redshift.And,asyoucansee,theobservationsprettymuchfollowtheband.So,thereisn't
very much, in terms of color evolution, for the Hubble-type sequence galaxies all the way
afterredshiftofunity.
[slide3]Hubblespacetelescopehasalsoaffordedusanotherpossiblemeasurement,which
is to look at galaxy sizes, say, looking at their half-light radii or some other form of
objectively defined radius. And so, this is the result of what the radii, average radii of
galaxies,doasafunctionofredshift,namely,galaxiesgrowintime.Iftheradiiwerefixedin
propercoordinates-iftherewasnoevolutionatall-thensolidlineshowswhatfwouldlook
like.But,insteadofthat,thepointsshowthatthereisasubstantialgrowthofgalaxies.This
isconsistentwithourunderstandingofhowdiskgalaxiesseemtoformfromtheinsideout.
Asyou'llrecall,thebulgeistheoldestpartinthemiddle.Then,youhavestellardisk.Then,
hydrogen extending beyond it. That hydrogen gas has to turn into stars. And, here we'll
probablyseethecollectiveeffectofthat.
[slide 4] Modern spectra of high-redshift galaxies can be used to measure not just the
redshifts,butalsovelocitybroadening,inotherwords,getavelocitydispersion.Withradii
measured from Hubble space telescope, we can infer dynamical masses, or from fitting of
thespectralenergydistribution,wecaninfertheirstellarmasses.So,thatisshownhere.The
sizesofsymbolscorrespondtothemagnitudes.Wecanseethatthemostmassivegalaxies
seem to be already in place at redshift 1 or 2 – this is sort of the upper envelope of this
distribution–whereasthelowermassesseemkepttobeevolving–thisisknownasgalaxy
downsizing. Naively, you would expect a hierarchical formation scenario, that you make
smallonesfirstandthenyougraduallybuildthebigones.But,thatseemstobetheopposite
to what's observed. The solution to this is probably due to biasing effect we talked about
earlier.
[slide5]Wecannotwaitlongenoughtoseegalaxiesmerge,butwecanassumethatsome
numberofcloseprojectedpairswilleventuallymerge.So,bydoingstatisticsofclosepairsof
galaxies,allowingforprojectioneffectsandthingslikethat,wecaninferthelikelymerger
rateasafunctionofredshift.Andhereitis.It'sapowerof1+z,andit'smoreorlessexactly
what'sexpectedfromthemodernmodelsofhierarchicalstructureformation.
[slide 6] You will recall that when we talked about scaling relations, like the fundamental
plane,Imentionedthattheycanbeusedasasharpprobesofgalaxyevolution.Luminosity
function is a very broad distribution, but these correlations are, by construction, the
sharpest we can have. And so, if we can follow, say, the evolution of their intercept or
maybe slope as a function of redshift, we can gain new understanding in the evolution of
galaxies-inthiscase,ellipticals.What'sshownhere,intheopenpointsthroughwhichthe
lineisfitted,istheessentiallyzeroredshift,thefundamentalplaneedge-on.Thesoliddots
areellipticalsinoneofthemostdistantclusternowknown.Therearetoofewpoints,but
youcanseethatthey'reconsistentwithbeingonashiftedversionofthefundamentalplane.
[slide7]Numerousstudieshavebeennowdone,bothforfieldandclusterellipticals.And,it
was always seen that ellipticals at higher redshifts show a shift in the intercept to the
fundamentalplane–thesurfacebrightness–thatwouldcorrespondtothefadingofstellar
populationintime,asexpectedfromevolutionarystellarpopulations.And,wecanevensay
somethingaboutthestar-formationhistories.
[slide8]Wecanfitdifferentevolutionmodels,andfindoutwhichonesseemtodescribethe
datathebest.And,theansweristhatmodelsinwhichellipticalgalaxiesformrelativelyearly
onanddon'tevolveverymuchsincethen,justpassivelyfadeawayseemtofitthedatafairly
well.Thatisshownintheplotontheleft.Theplotontherightshowsfadingor,ifyoulook
with redshift, brightening of surface brightness, which, remember, is related to luminosity
density of stellar populations as a function of redshit. And, you can see that it is
systematicallyincreasinginredshiftforbothfieldandclusterellipticals.
[slide 9] So, these studies are consistent with what we've seen earlier, that as far Hubble
sequencegalaxiesareconcerned,thereseemtobeprettymuchinplacebyaboutredshitof
unity.Theellipticalsareevolvinginawaywewillexpectfromearlyburstofstarformation,
andrelativelymodeststar-formationhistoryafterthat.And,wealsobegintoseechanging
thetiltoffundamentalplane,whichreallymeansthatgalaxiesofdifferentmassesevolveat
slightly different pace. And, it goes in the sense that those at the lowest mass end evolve
fastest,whichis,again,anotherexampleofthegalaxydownsizing.
That's at first fundamental plane's concern. What about Tully-Fisher? Well, that has been
doneaswell,butitturnsouttobemuchmoredifficultandcomplextodoitforspiralsat
highredshifts,andtheseresultsarestillnot100%clear,butareatleastbroadlyconsistent
withthispicture.
[slide10]Sofarwe'velookedattheevolutionofgalaxiesinthefield.Whataboutclusters?
Thisisadenseenvironment-youexpectgalaxyinteractionstoplaysomerole.Thefirsthint
that something interesting is going on in clusters was so-called Butcher-Oemler effect,
established very early on. These astronomers found out that clusters at larger redshifts
seemstocontainlargerproportionofbluergalaxies,forwhateverreason.Generically,you
expect the galaxy evolution leads from bluer galaxies to redder galaxies, so, at least
qualitatively,thisseemedaboutright.
[slide 11] Remember that in clusters some merges will occur, but vast majority of
interactions will not lead to merging. However it will disrupt galaxies, may remove tidally
some of their gas or stars and dark matter, and that's the process called the galaxy
harassment.There'llbealotofcumulativesmallencountersthatwouldmayberemovegas
from galaxies little by little. This would tend to transform late-type spirals and dwarf
irregularsintoS0sandellipticalsintime.
[slide 12] With the Hubble space telescope it became possible to look directly at a
morphologyofgalaxiesindistantclusters,andhereisoneofthem,wheredifferentsymbols
correspond to galaxies of different morphological types. With that and spectroscopic
measurementswecantrytodisentanglewhat'sgoingon.
[slide13]Aninterestingfindingwasmade.Thereisanoveltypeofgalaxiesfoundinthese
evolving cluster populations, so-called post-starburst galaxies, galaxies that are bluer than
theyshouldbeatthezeroredshift-therewasnothinghappening-andhavespectralenergy
distributions consistent with having undergone a burst of star formation maybe up to a
billion years earlier. This could have been caused by some interactions, and these galaxies
willthenpresumablyfadeintoredderHubbletypes.
[slide 14] And so, the summary of these results is that, as far as we can tell, there is a
conversionoflate-typespiralsintoS0sandellipticalsintheclusters.Thisisrelatedtodensity
of clusters as well – the morphology-density relation – and that can account for the
observedButcher-OemlereffectandpresenceofthingslikeE+Aorpost-starburstgalaxies.
[slide15]Soonepossiblescenarioisthatspiralsfromfieldarefallingintoclusterpotential
well. As they do that, they encounter dense intracluster gas, the extra gas. Some of their
owngasgetsstrippedaway-thatquenchesthestarformationinthem.Theymayundergoa
burstofstarformation,triggeredbysomeoftheinteractions,buteventuallythatfadestoo.
Andintheendyoumayhaveanolddisk,likeS0galaxy.So,thisisaplausiblescenario.Thisis
not necessarily the only way to make 0s, but it can account for the observed effects in
clusters.
[slide 16] Next, we will talk about the obscured component of star formation in history in
galaxies.
13.5DustObscuredGalaxies
[slide1]Sofar,welookedattheobservationsinvisiblelight,whichalsocorrespondstothe
redshiftedrest-frameultravioletlightingalaxies.Inotherwords,whatyoujustseedirectly
instellarsurfaces.However,wedoknowthatgalaxiescontaindust.And,likelytheydidso
fromveryearlyon.Thedustabsorbsultravioletandvisibleradiation,andre-emitsitinfarinfrared or submilimeter. Thus, there has to be a obscured star-formation component
involvedingalaxyevolution.
[slide2]And,indeed,asweaquiredabilitytoobservedeepuniverseonthesewavelengths,
itwasdiscoveredthattherearenewsourcesappearingintheskythataresimplynotvisible
at all in visual light. Here we see a composite in the upper left, visible light image from
Hubbleandthemid-infraredimagefromtheSpitzerSpaceTelescope.Asyoucanseeinthe
upper right, there isn't a trace of that red source. In the lower left, you can see it is just
beginningtoshowup,zerotonear-infrared.And,inthelowerright,it'spuremid-infrared
image,andsourceisveryobvious.So,skywouldlookdifferentonthesewavelengths,and
therecouldwellbeapreviouslyunaccountedpopulationofsources.
[slide 3] And this is, indeed, the case. The first observations that uncovered such a
population of sources were done at James Clerk Maxwell telescope in Hawaii. It is a
submillimeter telescope. It is equipped with a bolometer array measuring submillimeter
radiationfromsourcesinthesky.Theseareverydifficultobservationstomake;thisiswhyit
tooksolong.And,theylookedatacoupleofthedeepfields–theHubbleDeepFieldand
another one called Lockman Hole – and found that there are sources there, faint sources,
whose nature at the time was unknown but were of surely obscured star-forming galaxies
faraway.Thereasonwhytheylooksobloppyisthepoorresolutionofthesetelescopes.You
mayknowthattheangularresolutionoftelescopeisproportionaltoitsdiameterdividedby
thewavelength.So,foroptical,thisisaveryhighnumber.Therearemanywavelengthsof
photons stretched over, say, the mirror of the Keck telescope. Not so in radial or
submillimeter – there you have very low resolution. Nowadays, we have new
interferometers, like ALMA in Chile, that will actually produce optical light resolution in
thesewavelengths.Butbackthen,thiswasn'tthecase.
[slide 4] Now, here is a really nearby example of what we might expect. This is the galaxy
M82, which is a nearby starburst galaxy. The picture shown here combines regular visible
light and the purple is ionized hydrogen emission. What we see here is an intensely starforming disk galaxy obscured in the middle, but the supernovae exploding pushed the gas
out.Theydriveagalacticwind,expellingitintointergalacticspace.Now,ifwetakeabroadbandspectrumofM82,weseethattherearetwobumps.Thereisone,theoptical–which
correspondstoasortofquasi-blackbody,sumofallstellarphotospheres–andthebigger
oneinfar-infrared–whichisthermalemissionfromdustthatwasheatedupbytheseyoung
stars. In this particular galaxy, there is actually more energy emerging in the form of
submillimeterorfar-infraredradiationthanvisiblelight.But,onaverage,inthispartofthe
universe, there is roughly equal amount of obscured and unobscured star formation. The
sameturnsouttobetrueathigherredshifts.
[slide 5] You may recall the concept of K-corrections, which means that as you observe in
some stationary instrument on planet Earth, you're looking at different parts of the restframespectrumforsourcesatdifferentredshifts.Now,byandlarge,forgalaxies,thismakes
themdimmer,becausegalaxyspectratendtoberedder,moreenergyintheredpartofthe
spectra. It turns out that for submillimeter sources it's exactly the opposite. You are
redshiftingintothePlanckcurve-you'reclimbingfromtheWienside.Andso,eventhough
sourcesgetfurtherawayandthereforeshouldbedimmer,you'resamplingabrighterpartof
theirintrinsicspectrum.Andso,someofthoseK-correctionsactuallybecomenegative.The
upshot is that for many of these sources the apparent brightness almost does not change
withredshift.And,therefore,ifyoucanreachthatfluxlevel,youcanseeveryfaraway.This
iswhatSCUBAsourcesturnouttobe.
[slide6]...whichalsomeansthatwecandofairlydeepsourcecountsinthesewavelengths,
and here are some examples of it. They can be used then to constrain directly the
contributionofthesesourcestotheoverallstar-formationhistoryintheuniverse,andsay
somethingabouttheirevolution.
[slide7]Thereisonedifficulty,however.Thepoorangularresolutionofthesesubmillimeter
observations means that optical counterparts, which we need in order to measure the
redshifts, are hard to guess. In some cases, there is an obvious counterpart, but, in many
cases,therearemanyfaintgalaxiesinsidethiserrorcircleofasubmillimetersource.It'snot
clearwhichone,ifanyofthem,istheactualcounterpart.So,ittookawhiletogetredshifts,
measured for these objects. The trick that was used is that the radio measurements can
detect some of them, and radio measurements have precision that is perfectly good to
matchtoopticalobservations.
[slide 7] Then the spectra were taken of those, and, in fact, may turn out to be strong
emission-line sources, where that line emission could be powered by star formation or,
possibly, by an active nucleus in those galaxies. Possibly, a little bit of both is happening.
And,thepredictionsfromfittingthesourcecountswerethatmanyofthesesourceswillbe
aroundredshifts2or3,andthatexactlyturnouttobethecase.
[slide 8] So, now that we've seen how we can detect both obscured and unobscured
component of evolving galaxies, we'll look into the overall star-formation history of the
universe.
13.6:TheStarFormationHistoryoftheUniverseandtheChemicalEvolution
ofGalaxies
[slide 1] So, now that we've seen all these measurements of evolving galaxy populations,
let'sseeifwecantieittogetherintheoverallstar-formationhistoryoftheuniverse.
[slide2]Hereisaplotthat'scalledaMadaudiagram,afterthefirstauthorofthepaperthat
introducedhim–PierreMadau.And,itshowstheinferredproductionofmetals,ordensity
of star-formation rate, as a function of redshift, as inferred from Hubble Deep Field
measurements.Thethreesetsofpointsandcurvescorrespondtothreedifferentfilters,and
you can see that they're in reasonable agreement. The actual curves are particular galaxy
evolutionmodelsthatarefittothedata.And,theinterestingthinghereisthatyoucansee
that there is a peak. There is a time in the history of universe, roughly around redshift of
unity,whereitseemsthatco-movingstar-formationrateoverallgalaxieswasatitshighest.
And,itdeclinedsincethenasitapproachedredshiftofzero.Italsoseemedtohavedeclined
athigherredshifts,whichyoumightexepectwouldbethecaseaswefirstbuildupgalaxies,
they get brighter and brighter and then they fade off. Turns out, a lot of this apparent
decreaseofhigherredshiftsisactaullyduetotheobscuration.But,let'shavealook.
[slide3]Now,wecanmeasurereddeningsofgalaxies,andwecancorrectfortheapparent
absorptionbydust.And,ifyoudothat,youplotthesamediagram-thiswouldbediagram
ontheupperrighthere,duetoSteidelandcollaborators.Youfindoutthat,trueenough,as
yougofromheretotheredshiftof1,theco-movingstarformationdensityoverallgalaxies
increasesrapidly,butthen,itkindofstaysflatallthewayouttoredshiftof4,andmaybe
even higher. Now, plotting against the redshift is maybe slightly misleading, because you
really want to plot as a function of the lookback time – time history of the universe. And,
whenyoudothis,thenthisdeclineafterredshiftofunityisactuallymuchmoregentle,but,
nevertheless,thereisadecline.
[slide4]Thatwasallfortheunobscuredstarformation,orjustmildlyobscured.Now,ifwe
addcomponentfromdusty,obscuredsources,liketheSCUBAsources,thenthatboostsup
thetotalevenhigher.Thecurvesshownherearemodelsthatdon'tseemtofitterriblywell,
but they do imply that there'll be substantial component of hidden star formation
contributingtotheoverallhistory.
[slide 5] Subsequently, measurements have been obtained out redshift of 6, where we
actually have spectroscopic measurements of galaxies, and nowadays they push them to
redshiftoftheorderof10,butthoseareallbasedonphotometricredshifts.Andyes,there
is a decline that sets in roughly past redshift of 4 or 5, and that's the picture that we
expected.Atfirstthereisnothing,thenyoubuildupgalaxies,yougetmoreandmorestarformationrategoingon.Thereisaverybroadmaximumofthataroundredshift2+/-factor
of2,andthenitdeclinessincethen.So,wearenowinthephaseofhistoryoftheuniverse
where it's gradually fading away. Most of the action happened when universe was a few
billionyearsold.
[slide 6] We can convert these measurements into the actual buildup of the present
observedstellarmassingalaxies.And,hereiswhatitlookslike.Itstartsathighredshiftsasa
verysmallfraction.Byaboutredshiftofunity,mostofthestellarmassingalaxiesisalready
assembled, and then stays roughly constant. This is very much consistent with everything
elsewe'veseenbefore.
[slide 7] The more modern observations push this further out to redshift of 6, now even
beyond,andthetrendcontinues.So,westartwithnogalaxieswhatsoeverinourstars,and
buildthemupgradually.Therateofthebuildupslowsdramaticallyaroundredshiftofunity.
And,thereisstillsomebuild-upbecausethereisstillsomestarformation,butwenowcan
actuallyseegalaxiesbeingassembled,starsbeinggenerated,andgalaxiesovercosmictime.
[slide8]So,thiswasthedirectapproachtolookingatgalaxyevolution.Welookatindividual
sources,measuretheirdistances,brightness,andsoon.Analternativewayistoobservesky
and integrate all of the energy that we get from it, and actually obtain spectrum of that
overallintegratedcosmicbackground.Note,thisisnotthecosmicmicrowavebackground.
Thisistheradiationduetogalaxies,starsandgalaxies,andalso,maybe,activenuclei.
Thisisaverydifficultmeasurementtodo,becauseit'shardtogetzerocomparison.You're
lookingatthenormalpatchoflightandcomparingitto–what?So,thisiswhyittookso
long to do it. But, nevertheless, it was done in both optical and near-infrared. And, the
upshotisthatthetotalintegrateddensityofopticalandinfraredbackgrounds,almostallof
whichisduetostarformation,isabout100nanowattspermetersquaredpersecond.Ifyou
have telescope of certain collecting area, and integrate some time, that will give you the
amountofpowercollected.Thisturnsouttobeonlyfewpercentoftheenergydensityof
cosmicmicrowavebackground.And,ofitonlyfewpercentisactuallycontributedtoactive
galacticnuclei.
[slide 9] So, here is the broadband picture. This is what happens when you integrate the
spectrumoftheuniverse,inopticalandinfrared.There'redifferentmeasurements;there're
upper limits. They come from ground, from space, a variety of sources. If you recall the
broadbandspectrumofStarburstGalaxyM82,ithadtwohumps.Therewasblack-bodyish
radiationfromunobscuredstars,inthevisiblelight,andtherewasblack-bodyishradiation
fromheateddustinfar-infrared.Well,thesamenowappliestotheoverallspectrumofthe
universe,ifyouwill.And,what'splottedhereisaquantitythat'sreallyproportionaltothe
energy per unit logarithmic interval. And, the fact that a two peaks are roughly the same
heightmeansthatapproximatelyequalamountofallstarformationeverwasinunobscured
andobscuredsystems.
[slide 10] Let us now turn to the question of chemical evolution. As stars are made, they
explode, they release chemical elements and interstellar medium, new stars are formed.
Someofthoseareexpelledinintergalacticgas,freshgascomesin.Allofthatcontributesto
theoverallchemicalevolutionofgalaxies.How'stheirmetallicitychangingasafunctionof
time?
Hereisaroughschematicdiagram.Youbegin,ofcourse,withhydrogen,heliumandnothing
else. So, stars cook up heavy elements. Some of those recycle into new stars, some are
expelled out, fresh material comes in, and so on. Often times, these extremely complex
processesaresimplified.Forexample,galaxiesareputallinabox,andthereisnoinflowor
outflow. Or, it's assumed that the moment stars explode, that material is immediately
recycledinnewstars.Thosearecrudeapproximations,buttheygiveusatleastsomeinsight
ofwhat'sgoingon.
[slide11]Anotherschematicdiagramwhichconveysmoreorlessthesamestoryisshown
here. But, you may want to look at it and find it a little more informative as to all the
differentconnectionsbetweendifferentprocessesandcomponentsare.
[slide12]Now,starbusts,likethatoneinM82,candrivegalacticwinds,thatexpelenriched
materialjustproducedbysupernovae,thankstothekineticenergyofsupernovaejectaout
intheintergalacticspace.Thiscanbemodeledandisalsoobserved.Whathappenstothat
enrichedgasisthatitcontributestotheoverallchemicalevolutionofintergalacticmedium.
And,athighredshiftsitwillbeobservableintheformofmetalabsorption-lineclouds,which
we'lladdressinthenextchapter.
[slide 13] So, as you make stars, you make metals. The star-formation history of the
universe, and the chemical enrichment history of the universe are tightly coupled, and,
qualitatively, should look the same. So, here is, essentially, a Madau plot, that shows the
productionofmetalsasafunctionofredshift.
[slide 14] And, with modern measurement we can look at the dependence of galaxy
metallicity and mass. You'd expect that more massive galaxies would achieve higher
metallicities because they retain more of the supernova ejecta and recycle them more
effectively.And,thereis,indeed,amass-metallicityrelation.It'saverynoisyone.It'sshown
inthisplotasthegreyareathat'sfor,essentially,redshiftof0.And,now,thereisasetof
measurements for distant galaxies, now put in bins to make it more obvious. And, we see
thatthatkindofarelationshipexistsalreadyearlyon,whichiswhatyouexpect.Regardless
oftheredshift,moremassivegalaxieswillretainmoreoftheirprocessedmaterial.But,the
overall curve is shifted down, which is again what you expect – that you have lower
metallicities and they grow. They grow in galaxies of all different masses and at every
redshiftthereisthisdependencethatisgenerallyexpected.
[slide 15] ...which will lead us into the next chapter – evolution and structure of the
intergalacticmedium.