Observations of Galaxies and Large Sale Struture at

High Redshifts

Charles C. Steidel

Palomar Observatory, California Institute of Tehnology, Pasadena, CA USA

Abstrat

The last few years have seen rapid progress in the olletion of empirial data on

galaxies and their large-sale distribution at high (z >> 1) redshifts. Here we summarize

some of this progress, and disuss the level at whih the high redshift observations are

in agreement with the expetations of a general inationary universe dominated by old

dark matter.

1 Introdution

It is diult to know exatly where a disussion of suh a mundane topi as galaxy formation

ts in when we are hearing about muh more fundamental issues bearing on the status of

inationary osmology. Galaxy formation is an inherently \messy" proess whose onnetion

to the underlying osmologial model is indiret at best, and misleading at worst; trying to

understand it is really the lowliest form of \mud{wrestling" (to borrow the term from Martin

Rees, who introdued it as a ounterpoint to the more rened ativities that are the primary

fous of this symposium!). Nevertheless, galaxies are an observable onsequene of the universe

devised in the minds of many of the distinguished speakers at this meeting, and so I will

take it as my task to briey report on the extent to whih the observed universe of galaxies,

and their large-sale spatial distribution, is \as expeted" within the ontext of inationary

osmology. Put simply, we ask the question: do observations of early galaxies (here of ourse

early is a relative term!) support the general paradigm of the growth of matter utuations by

gravitational instability for a universe dominated by non-baryoni dark matter?

There has been a great deal of progress in just the last few years in obtaining empirial

information on the status of the universe at high redshifts (again, \high" by galaxy standards);

we now have observational a

ess to about 90{95% of the age of the universe, lling in muh

of the \gap" between deoupling at z 1100 and the present. We an now begin to follow the

progress of struture formation (as traed by galaxies) relatively soon after the rst galaxies

an have formed. These new apabilities oer the possibility of observing galaxy formation

diretly, and at the same time onfronting general \paradigms" for struture formation and

the omplex oupling between formation of galaxy{sized objets and the larger{sale matter

distribution.

In what follows, I will briey summarize this progress, and try to ast the interpretation

of these observations in a ontext that may be relevant to those interested in the broader

osmologial impliations of the galaxy formation proess.

2 Large Sale Struture vs. Galaxy Formation

2.1 Large Sale Struture Considerations

There are two prinipal ways of viewing galaxies in the ontext of observational osmology;

one way has traditionally been to treat galaxies as fundamental \test partiles" for exploring

the growth of matter utuations, and for measuring the matter utuation power spetrum.

In this view the atual properties of the galaxies are of little interest, and it is only their

distribution that matters. The galaxy properties neessarily enter in a pratial way in the

proess of the olletion of data; the spetral energy distribution and spetral features of a

galaxy have a signiant inuene over whether a redshift may be easily obtained, and the

luminosity enters in the sense that it limits the pratial size of a galaxy sample obtained for

the purpose of studying large sale struture. In fat, the size of the galaxy samples matter a

great deal; the universe is very lumpy on sales of at least 100 200 o-moving Mp, and

moreover many of the statistial questions one would like to answer about struture formation

an only be a

urately addressed with a huge sample of galaxies. The large sale struture of

the universe as traed by galaxies an in priniple be used as a diret test of the world model

through the shape of the power-spetrum (partiularly on sales of tens to hundreds of Mp,

whih is ruial in the ontext of most struture formation models whih involve old dark

matter), and possibly through the amplitude of the power spetrum given a

ess to reliable

estimates of higher order lustering statistis. These are among the main sienti drivers for

the grand galaxy surveys (e.g., the Sloan Digital Sky Survey and the Two-Degree Field redshift

survey being onduted at the Anglo-Australian Telesope) of the relatively loal universe being

undertaken as we speak (see Josh Frieman's artile in this volume).

The lear motivation for moving to higher redshifts is that it provides \snapshots" of the

progress of struture formation at muh earlier epohs. As suh, it oers the possibility of not

only haraterizing the struture in the universe, but also observing its evolution, following

it bak to an epoh where the universe is in many ways muh less ompliated (where linear

perturbation theory may still be appliable for most relevant sales) and ertainly loser to its

initial onditions. Naively, one might think that the evolution of galaxy lustering would be a

fairly lean test of the overall matter density

m

, as one learns early in life the standard lore that

the growth of linear matter perturbations essentially eases one

m

(z) departs signiantly

from 1.0. One might expet, then, that lustering evolution of galaxies as a funtion of redshift

would be very

m

{dependent, with muh stronger evolution expeted for an Einstein-de Sitter

universe than for a universe with

m

(z = 0) ' 0:2 0:3. This would ertainly be true if

galaxies were reliable traers of the growth of matter utuations. Unfortunately, galaxies

are themselves both forming and evolving at the same time that the matter utuations are

growing, so that the mapping of luminosity onto mass is expeted to hange in ompliated ways

as a funtion of time. This means that to use galaxies as traers of large sale struture over

signiant time baselines requires some insight into what onditions result in galaxy formation,

what physial proesses are dominating galaxy evolution.

2.2 Galaxy Formation Considerations

The subjet of galaxy formation, as opposed to the study of the development of large sale

struture, has a muh longer history, and therefore, to some extent, arries more historial

\baggage". Long before there existed muh information on \large sale struture", the so{

alled \fossil reord" of nearby galaxies, inluding the Milky Way, provided a piture of galaxy

evolution through the kinematis, ages, and hemial ontent of the onstituent stars. Broadly

speaking, the oldest stars in nearby galaxies tend to be parts of the spheroidal omponent

(elliptial galaxies, and the bulges/spheroids of spirals) and many onsiderations led to the

onlusion that these stellar omponents formed > 10 Gyr ago, and may well have formed on

roughly one dynamial timesale { a so{alled \monolithi ollapse" ([12, 46). This general

piture of rapid and early formation of spheroids prompted many searhes for \proto-galaxies"

at high redshift beginning in the 1960s and 1970s, and ontinuing to the present. Although it

has been appreiated for some time that galaxy morphologial types tend to depend on the loal

galaxy density, with the bulk of elliptial galaxies inhabiting the rihest environments, there

was little onnetion made to large{sale struture in the lassial piture of galaxy formation.

Galaxy formation was seen as an isolated event, in that the origin of the mass onentration

that proeeded to ollapse was not addressed, nor was the onnetion of that lump of matter

to the overall struture of the universe.

The \modern" paradigm for galaxy formation attempts to work in the other diretion.

That is, one begins with a spetrum of density utuations (usually assumed to be initially

Gaussian and often with a sale free spetrum as predited by ination) and tries to understand

galaxy formation as a natural onsequene of the growth of the utuations by gravitational

instability. In this piture ([51), galaxies ome into being when baryons ool within virialized

halos of dark matter, and eventually form stars and beome diretly observable. The mass

funtion of virialized objets is onstantly hanging due to gravitational instability over time,

and the abundane and spatial distribution of ollapsed objets of a given mass will be both

time and osmology dependent. The utuation power spetrum is tuned to produe the right

power on various observable sales (for example, one might insist that the power spetrum of

one's favorite model produe the right power on both COBE (super-horizon) sales and on

the sales of rih lusters of galaxies simultaneously{ galaxies would then fall out naturally on

the small-sale tail of that power spetrum), and one that is xed, one ought to have an ab

initio handle on the epoh of formation and the evolution of galaxy-sized lumps of matter as a

funtion of osmi time.

There are a number of ompliations, however; one of the more serious ones is that the

distribution of galaxies and the distribution of virialized halos is not the same and the nature of

the mapping will hange with redshift. For example, rih lusters of galaxies onstitute a single

dark matter halo with a mass of 10

15

M

, yet they ontain hundreds of individual galaxies all

of whih at the time of their birth presumably had their own distint dark matter halos. Beause

of the ombination of ontinuing star formation, dissipation of baryons, stellar evolution, and

the hierarhial merging of dark matter halos, galaxies in an observed sample (whih will

always depend on the way in whih objets were seleted, and how those seletion riteria

might inuene the ontents of the sample as a funtion of osmi epoh), it is not neessarily

the ase that observed galaxies at one epoh are the diret desendents or anteedents of those

observed at another. In general, the spatial distribution of virialized objets is not the same as

the dark matter distribution; this is partiularly true for the most massive (rarest) peaks in the

matter distribution at any osmi epoh (\high peaks bias"{ [23). Large N-body simulations

are now apable of following in detail the distribution of mass and of ollapsed halos versus

time, so that given the right osmologial model (power spetrum shape and normalization,

matter density, Hubble onstant,

, et.) one might hope to predit exatly where galaxies

should form, and when. These preditions would then be observationally testable.

Perhaps the most intratable limitation of omparing theory to observation at present,

though, is our ignorane of the details of the behavior of the baryons within the dark matter

halos. The proess of star formation is not well understood; to take proper a

ount of the

probability that the position of a partiular lump of matter with a partiular mass would be

observable, one needs to understand the rate of gas ooling to form new stars, the stellar

evolution of stars born prior to the epoh of observation, the inuene of energy feedbak (e.g.,

from supernovae, or from photoionization from the metagalati UV radiation bakground)

on subsequent generations of stars, and what happens to gas and stars when their parent

dark matter halos merge with others. The observability of a given bolometri luminosity also

depends sensitively on the detailed geometry and on the hemial abundane of the material{ as

disussed below, prodigious amounts of dust formed during episodes of star formation will aet

the extent to whih a galaxy is observable using a partiular tehnique. To date, the theoretial

approahes to these problems are either to add baryon physis \by hand" using semi-analyti

models and saling relations for star formation, energy feedbak, et. (with onstraints on model

parameters oming from observations), or alternatively, by adding hydrodynamis diretly to

the N-body simulations. Both of these tehniques have been quite su

essful in addressing

many of the observations disussed below (see, e.g., [15).

Thus, galaxy formation involves a huge number of physial proesses, ating over a large

range of physial sales. Galaxies an be though of as a non-linear, evolving map of the

dark matter distribution, and as suh, their use for osmology is highly suspet without well{

developed ideas about their nature. On the other hand, understanding galaxy formation is

losely tied to the struture formation paradigm that one is interested in testing{ understanding

galaxy formation will involve solving for both the bakground osmology (of primary interest to

most of the people at this meeting) and the relevant physis that ultimately result in the objets

whih some (mainly astronomers!) nd fasinating in their own right. For this reason one

annot possibly separate the \use" of galaxies for osmologial purposes and the understanding

of how galaxies form and evolve. The best way forward toward an ab initio understanding

of galaxy formation/evolution is to onstrain the available parameter spae with extensive

observations.

3 Galaxies at \Very High" Redshift

3.1 The Optial/UV View

Redshift surveys arried out in the early to mid 1990s ([10, 27, 13) shed onsiderable light

on the galaxies to z 1, when the universe was about 40% of its urrent age. A ommon

onlusion from many independent lines of investigation was that the largest and most massive

galaxies evolved very little during that period of time, at the same time that there has been

onsiderable evolution in the total \star formation ativity" as measured by the total UV

luminosity of galaxies per unit o-moving volume. The inreasing (with redshift) star formation

ativity appears to be onned mostly to smaller, less regular looking galaxies that happen to be

undergoing episodes of star formation. Thus, there is already evidene for dierential evolution

of galaxies that to zeroth order seems to depend on galaxy mass. Elliptial galaxies in rih

lusters are among the most stable, indiating partiularly early formation epohs. From the

point of view of those interested in galaxy evolution/formation per se, it is learly neessary to

reah beyond z 1 to probe the formation epoh of the most massive galaxies. From the point

of view of those interested in struture formation and osmology, higher redshifts are neessary

to reah osmi epohs where the dierenes between model preditions would diverge most;

sine all models are tailored to produe the z = 0 universe by z = 0, reahing the largest possible

redshifts ould oneivably allow the strongest disrimination among theoretial possibilities.

For pratial reasons disussed below, the large sale struture traed by galaxies has remained

largely unexplored for distant galaxies; in fat there are arguably no reliable measures of galaxy

lustering properties beyond redshifts of z 0:3.

The standard means of onduting redshift surveys of faint galaxies involves obtaining spe-

tra of all objets down to some presribed limiting ux limit in an observed bandpass, over a

small solid angle on the sky. This method allows one to take advantage of multi-objet spetro-

sopi apabilities of modern instruments (for the faintest galaxies one uses a foal plane mask

whih positions slitlets on top of several tens of galaxies within a eld of 5{10 ar minutes on the

sky) and will sample a large range of galaxy redshifts. To rst order, the fainter the ux limit,

the higher the redshift to whih the resulting sample extends. However, this stops being very

eient for piking up substantial numbers of the most distant galaxies at z

>

1 for several rea-

sons; rst, the dierential evolution of galaxies alluded to above means that at higher redshift,

most of the objets at a given faint ux level will be less luminous \foreground" galaxies and

not luminous and distant ones. Seondly, even intrinsially bright galaxies are very apparently

faint; at z 1, a galaxy of roughly Milky Way luminosity would be less than one-tenth the

brightness of the dark night sky at 7000

A and suh ux levels were just barely ahievable on

the largest ground{based telesopes until the mid 1990s. In addition, the typial spetrum of

a galaxy forming stars is quite \at", with a roughly onstant ux per unit frequeny at wave-

lengths where the sensitivity from the ground is highest (4000-8000

A) due to high detetor

eieny and low bakground. The spetrosopi features whih enable easy redshift identi-

ations are rest{frame optial lines that ome from reombination in regions photo-ionized by

hot young stars. However, when z

>

1, these features have been shifted into the near-IR, where

the terrestrial bakground is muh higher. At rest-frame optial wavelengths, one sees only a

relatively featureless UV ontinuum so that unambiguous measurements of redshift in low S/N

spetra is very diult.

The UV ontinuum of young galaxies forming stars remains relatively featureless until the

Lyman edge of hydrogen, at 912

A(13.6eV, or 1 Rydberg). At this point any galaxy that is

radiating in the UV at all is guaranteed to have a pronouned disontinuity, as illustrated in

Figure 1. This feature is onvenient for a number of reasons: rst, it is so dramati that is does

not require either high spetral resolution or high S/N to identify. Seond, it is observable from

the ground, at near-UV/optial/near-IR wavelengths, for 2:5

1; the

hange in the behavior at the highest observed redshifts is almost entirely attributable to a

redution in the sample variane that plagues a small eld like the HDF, replaing those data

with the the data from wide{eld ground{based surveys at z 3 and z 4. The z 4

sample omes from a survey designed to establish the Lyman break galaxy properties in the

redshift interval 3:9