Abstract

Contents Updated: Wednesday, April 19, 2000

M Fukugita, C J Hogan & P J E Peebles answer the call of Chiu in 1968. Phew!

According to the relativistic Big Bang model, the Universe started with a uniform mass distribution, but was gravitationally unstable to the growth of clustering. The present concentrations of matter in galaxies and systems of galaxies grow out of small departures from homogeneity present in the very early Universe. Galaxy formation is much too complicated for a detailed analysis from first principles, so we attempt to piece together a picture of what happened from the observational evidence. For many years, the main elements of the picture depended on the fossil evidence in relatively nearby galaxies, including our own, but recent advances in astronomical imaging and spectroscopy have added a new dimension to the search for an understanding of how galaxies formed. We can see what happened over a considerable span of cosmic time. Observations now penetrate to such great distances that the light travel time from the source of the radiation to the present is comparable to the expansion time in the Big Bang cosmology, allowing a direct view of cosmic evolution from protogalactic gas clouds, to young galaxies, to the mature systems of the present epoch.

The light from distant objects carries information in spectra as well as images and from absorption processes as well as emission. The most direct visualization comes from images, the deepest now being the new Hubble Deep Field . These images show starlight in emission, allowing measurements of brightness, broadband colour and morphology, and conveying qualitative information about the amount of star formation and spatial organization in early generations of galaxies. Dispersion of light into 1,000 or more colours in the images of brighter galaxies allows a more detailed study of the emitting stars and gas, and especially, through the measurement of redshift, their distances. A few extremely distant quasars—the rare, bright, active nuclear regions of some early galaxies—are so luminous that their light can be dispersed into over 30,000 colours. At this resolution, line absorption by foreground gas reveals a detailed one-dimensional map of galactic and protogalactic gas all the way back along the line of sight to the quasar, through the last three-quarters or more of the history of the Universe

The directly measurable coordinate along the line of sight is not time but redshift (z ); the expansion of the Universe stretches the wavelength of the light from a distant galaxy by the factor 1 + z if the Universe has expanded by the factor 1 + z since the light we detect left the galaxy. Thus, measurements of the redshift date the slice of cosmic history that we explore by looking out in space and back in time along our past light cone-like the history recorded in an ice core, a tree-ring sample, or geological strata surrounding a dinosaur footprint, except that the astronomers’ record is imprinted on the images and spectra of light arriving from distant objects. The character of the emission spectrum contains information on the source of the light, whether from a young or an old star population. The star formation rates in individual galaxies can, for example, be estimated front the fluorescent line emission from gas excited by light from hot young stars or the shape of the ultra-violet continuum. In the case of quasar absorption spectra, the light from the quasar has passed through many different gas clouds at different redshifts, and each cioud has left in the observed spectrum an imprint of absorption lines at wavelengths determined by the redslhift of the cloud. The absorption lines from various ions carry information about the internal velocity, temperature, column density, composition and ionization state of the intervening gas.

Our purpose here is to review the progress in these types of observation and propose a synthesis of the results in a picture that contains many of the trends one would look for in a Universe that is evolving from a dense, nearly homogeneous early state—the clearing of intergalactic material as the gas is concentrated into galaxies, the emission from large numbers of newly formed stars in young galaxies, the change in composition of material from a nearly primordial mix dominated by hydrogen and helium to one enriched in heavier elements, the growth of galaxy clustering with time, and the changes in galaxy morphology as these objects approach their present slowly evolving state. We trace the evolution back in time from the present to what seems to be the epoch of formation of the giant galaxies, in the redshift interval z = 1 - 3 where both imaging and spectroscopy now yield abundant detailed information

Galaxies at z  <= 1

Giant galaxies today—those with luminosity similar to our own Galaxy—usually evolve only slowly, on time-scales comparable to the expansion time of the Universe (~10 Gyr), although there are examples of rapid evolution. Present-day galaxies serve as the baseline for comparison to the younger-looking, more rapidly evolving objects observed at redshifts greater than unity.

Giant galaxies are often divided into two major categories, spiral and elliptical The visible light of spiral galaxies comes mostly from disks of stars on nearly circular orbits. Stars are forming from gas in the disks, but the rate is usually slow because the mass fraction in gas is low and star formation tends to be self limiting—formation of stars heats the gas and prevents further small-scale gravitational collapse. Elliptical galaxies contain almost no neutral or molecular gas. The ageing stars in an elliptical galaxy orbit more isotropically than the younger stars in a spiral, filling a triaxial (roughly ellipsoidal) shape. Ellipticals are denser than spirals, and tend to be found in denser environments. These simple categories are not sharply defined, however—spiral galaxies often have an elliptical-like spheroid component of old stars, ellipticals sometimes contain clouds ofgas and dust and young star clusters, and the intermediate S0 galaxies look like spirals that have been stripped of gas.

In the high-angular-resolution images frorn the Hubble Space Telescope (HST) the giant galaxies at z  ~ 1, both ellipticals and spirals, are somewhat bluer than their present-day counterparts, which is evidence of greater star formation rates, and spiral disks look a little fuzzier, yet to be settled into the prototypical morphology, but they otherwise look much the same as they do today. Consistent evidence comes from a deep redshift survey of galaxies selected in the near-infrared (the I band centred on wavelength 0.85 μm). The detected light is emitted in the blue (taking account of the redshift), and probably comes from a mix of stars similar to what is seen in present-day normal galaxies. The counts of galaxies as a function of brightness and redshift are about consistent with what would be expected from the properties of the nearby sample of galaxies. The lack of rapid evolution to z  ~ 1 or even higher also is indicated in the deep galaxy counts in the infrared K band (2.2μ-m)and in the giant galaxies identified from quasar absorption line studies. That is, the evidence is that the giant galaxies already were mature slowly evolving systems at z  = 1.

In certain circumstances, these giant galaxies evolve faster. The equilibrium in the disk is disrupted on occasion by interactions between galaxies. The tidal interaction triggers flows of gas and angular momentum, leading to rapid conversion of gas to stars at rates well above the sustainable equilibrium, often resulting in starbuist ga}axies. Such galaxies are usually very bright in infrared emission from dust produced along with the burst of young stars. The dust absorbs and reradiates most of the light from the massive young stars. The energy deposited by the stars in the interstellar gas may eventually quench the starburst and leave some diffuse gas in the galaxy, or it may drive most of the remaining gas out of the system.

Even in the Universe today there are examples of the central process of galaxy formation, the exchange of gas between galaxies and the intergalactic medium. The sweeping streams in the distribution of the gas around the central three galaxies in the nearby MSI group suggest the gas was swept out of the galaxies, perhaps mixing with a component falling in from greater distances. The stars in the galaxies are likely sources for the heavy elements in these gaseous haloes, and the halo gas may be falling back into the galaxies and contributing to star formation. But neither starbursts, nor the exchange of gas with the intergalactic medium, can have had much effect on the evolution of most giant galaxies at redshifts z < 1, because these objects show so little difference from today’s galaxies.

The slow evolution of the giants is not the whole story. Gas-rich galaxies, particularly the less luminous ones, were much more numerous than today even at modest redshifts. There were two early indications of substantial evolution at modest redshift—rich clusters at z  ~ 0.5 contain a larger proportion of gaseous star-forming systems than is observed nearby, and counts of galaxies selected in blue light (that is, light emitted in the ultraviolet) are larger than would be expected if the galaxy population were not evolving.

There is no generally accepted interpretation of all the details of the deep galaxy counts, but it is known that there are significant contributions to the excess blue counts from galaxies that are somewhat less luminous and bluer than is typical of the giants, and more irregular in form, as seen in HST images. The colours, irregular forms and decreasing abundance with decreasing redshift suggest that we are seeing galaxies made bright in the ultraviolet by massive short- ived stars that are rapidly exhausting the supply of gas for new generations of stars. This suggests that star formation in these galaxies has started only at: z < 1.

This picture of delayed star formation in less-luminous galaxies is tested by studies of the spread of ages of stars in present-day galaxies. There is evidence that the stars in the halo of the lilky Way galaxy have a small scatter in ages, as would be expected if the spheroid in our Galaxy formed along with the elliptical galaxies at high redshift. The distribution of star ages in nearby extreme dwarf galaxies is seen to be broader, consistent with late and non-coeval formation and evolution in lower-mass systems.

Diffuse Gas at z  < 1

We have a remarkably detailed picture of the distribution of gas clouds outside the galaxies from the spectra of quasars. Clouds that happen to lie along the lines of sight produce absorption lines at wavelengths determined by the cloud redshift or distance. This gives us an unbiased census of diffuse atomic hydrogen and heavier elements wherever the gas maybe.

At redshifts z <~ 1 gas clouds with relatively high column density Σ_H1 = 3 x 10¹⁷ to 3 x 10²¹ neutral atomic hydrogen atoms per cm²) invariably are in the vicinity of a galaxy, at a radius that depends on the cloud column density, the atom and its ionization. At z  ~ 1, most giant galaxies have the halo of absorbing clouds shown schematically in Fig. 2. For example, clouds with column density Σ_H1 = 10¹⁷cm^-2 (the so-called Lyman limit at which the gas is just optically thick to ionizing radiation at the threshold) are found only within about 40 kpc of a giant galaxy. Apart from a weak dependence on mass, this absorption profile is independent of almost all properties of the galaxy, suggesting a connection with infall rather than stellar-driven outflow, and is remarkably close to spherical in character, suggesting a quasi-static, pressure-regulated process rather than a free ballistic infall. The gas clouds shown in Fig. 1 have column density Σ_H1 >~ 10²⁰ cm^-2 (approx), which is well above the Lyman limit, and are larger in extent than is typical of present-day galaxies, but they illustrate the point that gas clouds near the Lyman limit tend to be very near galaxies.

At lower column densities, Σ_H1 ~ 10¹⁴ cm^-2, the clouds still tend to be within 200 kpc or so of a galaxy. That is, most of these clouds populate regions which occupy only one part in 10³ of the volume of space. At the present epoch, and at the lowest observable column densities, Σ_H1 ~ 10^l2-10^l3 cm^-2, the gas clouds avoid the immediate neighbourhoods of galaxies, but they also tend to avoid the broad empty regions between the concentrations of galaxies.

In summary, at the present epoch and back to redshift z   ~ 1, baryons in clouds leftover from galaxy assembly have been quite efficiently swept from the voids between the concentrations of galaxies, leaving the galaxies with gaseous haloes that may be contributing to the present generally leisurely rate of star formation.

Mean Densities of Baryons in Stars and Gas

Estimates of the amounts of baryons (the neutrons and protons of familiar matter) present in various forms in the Universe today are useful for comparison to the amount observed in the more rapidly evolving systems at higher redshift. There is evidence that we have a nearly complete survey of where the baryons are—the amounts seen in stars, cool gas clouds, and hot clouds of plasma add up to about what is needed for the theory of light-element production in the early hot Universe.

It is convenient to express the baryonic mass in form i as the ratio Ω_i, of the mean mass density in this form to the critical total mass density in a universe that is expanding just at escape velocity. (We use the distance scale parameter h ~ 0.7, where the Hubble constant is H₀ = 100h km s^-1 Mpc^-1. The uncertainty in this parameter is not the main contributor to the uncertainties in our estimates of the Ω_i.)

Table 1 lists estimates of the density parameters Ω_i. The star masses in spheroids (elliptical galaxies and bulges of spiral galaxies), disks and iiregular galaxies are estimated from models of the stellar populations. The relatively little neutral gas in the Universe today is found almost entirely in the disks of spiral galaxies. Atomic neutral gas contributes Ω_gash ~ (2.3±0.6) x 10^-4, and molecular hydrogen a similar amount. Hot plasma dominates the observed baryon content of rich clusters of galaxies (gas/total ~ 0.05h^-3/2)and is detected also in some group.

Most giant galaxies occur in groups which can harbour a significant number of "dark" baryons, either in the form of plasma or compact baryonic objects. Plasma temperatures in groups are lower than in clusters, consistent with the fact that the galaxies move more slowly, and consequently the plasma in many groups may be too cool to be detected directly as a source of X-rays. Compact baryonic objects (MACHOs) include planet-sized systems like Jupiter, stars with masses near or below the limit for sustained nuclear burning, and star remnants. Their abundance is now being probed in our Galaxy by microlensing searches. We have entered a number in Table 1 which assumes that the number of baryons per spheroid star mass is universal, the same in groups as in the rich clusters, as would be expected if most spheroids formed early, when conditions everywhere were still much the same. In Table 1, the smaller number for baryons in cool low surface-density clouds is more plausible if the clouds tend to be wispy rather than equally wide in all dimensions.

We have a measure of the total baryon density, Ω_b, from the relative amounts of the light elements produced in the hot young Universe. Acceptable fits to the light-element abundances follow if the baryon density is in the range 0.01 <~ Ω_b2 <~ 0.02. There is, however, an argument based on the high deuterium abundance found in some quasar absorbers that Ω_b is as low as 0.005h², or Ω_b ~ 0.01 if h ~ 0.7.

The lower number, Ω_b ~ 0.01, agrees with the total in the components listed in Table l,which allows room for dark baryons only in an amount comparable to that in each of the other major components, perhaps Ω ~ 0.003. If the total baryon density is at the high end of the range of estimates from light-element production, Ω_b = 0.04, the dark baryons must dominate over visible stars by an order to magnitude, which ceitainly is possible but requires special arrangement. In particular, if mass were sequestered in dark star remnants one would look for considerable debris from the star deaths. The debris could be in smoothly distributed plasma in the voids, but gravity probably swept the voids clean of plasma as well as galaxies and gas clouds. We do not need to invoke any mechanisrns to conceal baryons if total baryon density is Ω_b ~ 0.01. As we will discuss, this is comparable to the amount of baryons identified in gas clouds at larger redshifts.

Galaxies at Redshifts z  = 1 - 3

Although observations of galaxies and possible progenitors of galaxies at z  >~ 1 are usually biased towards objects whose activity makes them particularly noticeable, we do know that there are objects at high redshift that are quite unlike galaxies at the present epoch, and they are detected in numbers large enough that we must be seeing a major phase of star formation in galaxies. Thus a substantial part of the whole history of the galaxies is observationally accessible, at z <~ 3.

We now discuss some examples of what appear to be young galaxies and the gas clouds that are progenitors of the disks of spiral galaxies. Optical searches for the identification of radio sources have led to the discovery of young galaxies at redshifts above unity. The most distant galaxy discovered as a radio source is at redshift z = 4.25. The continuum radiation emitted at wave length 4,200Å (and detected at 2.2 μm) in this galaxy is thought to be starlight. If so, it is estimated from the colour that the stars already were about 0.5 Gyr old at the time of observation and that the star population would fade to about the luminosity of a large elliptical galaxy at the present epoch. The image in what was at emission ultraviolet light is aligned with the radio lobe axis in a way not commonly seen in low-redshift radio galaxies. This pattern repeats in other high-redshift galaxies discovered as radio sources. The images at optical wavelengths in emission tend to look like giant elliptical galaxies, while the alignment in images in ultraviolet light and in gaseous emission suggests the radio activity is still rearranging gas in and around the galaxy. At least one weak radio galaxy has been found at z = 1.55 with a spectroscopically measured stellar population age of 3,5 Gyr, implying a much earlier formation redshift.

Surveys of spectra of faint field galaxies reveal objects at z  > 1 with high star formation rates, as deduced from the intensitics of prominent gaseous emission lines, and irregular images in the optical in emission, which is suggestive of clumpy and very rapid star formation. A substantial subset of such star-forming galaxies, selected by examining colours for the tell-tale zero flux of starlight beyond the Lyman limit, extend beyond z  = 3. In most cases the luminosity profiles of these very high-redshift galaxies resemble modern elliptical galaxies (with half-light radii of only 1 or 2 kpc), but the flat spectral energy distribution is characteristic of a much younger stellar population. Indeed the spectroscopic properties (including interstellar absorption features. of heavy elements) resemble in remarkable detail those of the prototypical nearby starburst galaxy NGC4214, and the measured velocity dispersions (180-320 kms^-l) reproduce the range of modern giant ellipticals. The star formation rates in such galaxies, about ten solar masses per year, is enough to produce the number of stars in a giant galaxy if it continues for a Hubble time—a few billion years at that redshift. The number density of such star-forming galaxies is enough to account for substantial fraction of present-day giant elliptical galaxies.

Thus the evidence is that the giant ellipticals formed as galaxies of stars at redshift z = 3 or earlier. The close family resemblance of the spheroid components of spiral galaxies to ellipticals suggests the spheroid components also formed at z >~ 3 too. The thin disks of the spirals were added later.

A hint about the formation of the disks of spiral galaxies comes from the quasar spectrum absorption lines produced by gas clouds along the line of sight. The clouds with the largest column densities in neutral hydrogen, Σ_H1  ~ 10^21±1 cm^-2, are prominent because they display absorption in the broad radiation-damped wings of the Lyman-α transition, and hence are called damped Lyman-α clouds Wolfe has long emphasized that these clouds look very much like young galaxies, or gas-rich disks in the process of contracting to the present-day thin disks of stars in spiral galaxies.

The pattern of absorption-line redshifts of the low-ionization material in damped Lyman-α clouds is consistent with rotation in disks at circular velocities quite similar to the velocities in present-day spiral galaxies, even to z  = 4.4. The higher-ionization C_IV lines do not show the systematic rotational asymmetry, but the higher ionization suggests the C_IV is on the outskirts of the protogalaxy, where the matter is more exposed to intergalactic ionizing radiation. Thus it is reasonable to imagine that the C_IV is in clouds moving on disorganized orbits, The sizes and velocity dispersions inferred for the C_IV haloes are consistent with those of spiral galaxy haloes.

The total hydrogen mass in damped Lyman-α absorbers at z  ~ 3 is Ω_H1 ~ 0.0015-003h^-1 (or larger if dust in some clouds obscures the image of the quasar), comparable to what is seen now in stars (Table 1). The mass in the clouds decreases at lower redshift, as if the H_I were being consumed between z = 3 and z = 1.

The presence of heavy elements shows that stars have been forming and dying in these clouds even by z = 4.4. The heavy-element abundances at z = 2-3 tend to be about 10% of the solar abundance typical in the disk of our Galaxy, but with a spread of at least an order of magnitude either way. Some have been enriched to about solar while others are still relatively free of heavy elements at about 0.01 times solar. This large scatter is indicative of non-coeval disk formation taking place just around this redshift, because the heavy-element abundance in a closed system reaches the asymptotic value about 1 Gyr after star formation begins. The most enriched material is suitable for modern young stars: the gas would produce stars with the compositions in the thin disk stars of spirals. The less-enriched clouds may be contributing stars to the thick disk component and receiving heavy elements shed by stars in the thick disk and spheroid components. The pattern of element abundances in the clouds often (if not always) hints at that expected for enrichment by the first generation of stars, with relatively larger amounts of oxygen and silicon (produced by rapidly evolvolving stars) than carbon and iron (produced by slower ones), compared to the interstellar medium today. The same pattern is seen in old stars in our Galaxy.

Because the largest H_I column densities in absorption are about one-tenth of the typical accumulated total mass column densities in stars in the disks of present-day spirals, these damped Lyman-α clouds could not be converted to stars to form spiral disks without further rearrangement of the material. Some thin disk stars may already be in place in these clouds, and as gas is converted to stars more may be settling into the clouds. One such cloud at z  ~ 3 has been observed in a deep irnage that reveals the presence of stars. They may be early generations in the disk, or a spheroid component, or perhaps a close companion.

If the higher-column-density clouds at z  ~ 3 are associated with galaxies, then the cloud positions ought to be clustered in space, as are galaxies. The observed clustering is even stronger than in present-day galaxies.

To summarize, the evidence is that the disks of the giant spiral galaxies were assembled at redshifts in the ranget z = 3 - 1. We have a reasonable candidate for the disk progenitors, in the high-column-density damped Lyman-α clouds. There is evidence for rapid star formation at z  = 2 - 1, much of which we can ascribe to disk formation, and a corresponding decrease in thc neutral hydrogen mass in the damped Lyman-α clouds. The close resemblance of the spheroid components in spirals to elliptical galaxies leads us to suggest the spheroids formed with the ellipticals, at z >~ 3, the present-day disks being added later. The evidence for high circular velocities in the high-column-density clouds suggests the massive dark haloes of the spirals were attached to the spheroids before or along with the damped Lyman-α clouds. It is a suggestive coincidence that the space density of bright optical quasars also reached a peak sometime around z = 2-3. Perhaps thc high quasar activity is a by-product of the first assembly of giant galaxies, the addition of their disk components, and the accompanying massive starbursts.

Intergalactic Gas at Redshifts z  = 1 - 3

Intergalactic space between the concentrations of galaxies is quite thoroughly cleared of gas clouds at the present epoch. The situation is very different at high redshift: a foam-like network of low-suiface-density clouds fills space with mean baryonic mass density comparable to what then is in the damped Lyman-α absorbers.

The low-surface-density clouds, at neutral hydrogen column densities in the range 10¹³ <~ Σ_H1 <~ 10²¹ atoms cm^-2 appear in quasar spectra as a dense set of Lyman-αresonance absorption lines at redshifts less than that of the quasar—the so-called Lyman-alpha forest. The neutral hydrogen column densities in the forest clouds are too small to shield against the ionizing radiation from quasars and young stars, so the material is mostly ionized, with neutral fraction of about 10^-4 at the lowest columns, increasing in the higher-column clouds. At z = 3 the mass in plasma in the forest is comparable to the neutral mass in the damped Lyman-α absorbers identified as protodisks of spirals. Thus the total mass in the low and high column density absorbers at z = 3 is comparable to the total seen in all components at the present epoch (Table 1). The reasonable interpretation is that the two classes of absorbers at z = 3 are major contributors to the present- ay components listed in the table. Figure 3 sketches our impression of the evolution ofthe mass fractions Ω, in the major forms of baryons as functions of redshift.

Absorption by the forest gas in the equivalent resonance line of singly ionized helium also appears to be ubiquitous. The fact that a considerable fraction of the helium has been only singly ionized tells us the ionizing radiation does not have a very hard spectrum, especially at high redshifts—possibly due to the effect of gas absorption.

We have a coherence length for the mass distribution in the forest clouds from the comparison of spectra of quasars that are close together on the sky (either gravitationally lensed images of a single quasar or two different quasars). The forest absorption patterns are almost identical when the sightlines are closer than about 10 kpc, still similar at 100 kpc separation, and dissimilar at separations much larger than 100 kpc. This could mean either that the cloud sizes are about 100 kpc or that smaller clouds cluster on this scale (with a large two-dimensional covering factor), numerical models suggest that the coherence length is associated with the long dimensions of wispy clouds. The mean distance between clouds along a line of sight is 500 kpc to 1 Mpc, depending on the cosmological model As this is not much larger than the coherence length, the cloud network roughly fills space. Thus it is not surprising that forest clouds do not show strong clustering along the line of sight. There is not much room for the clouds to cluster.

There are heavy elements even in the forest material, The line strengths are consistent with an abundance about 1% of solar, and display little scatter among most clouds where it has bccn measured. This points to an early history of low-level star forma tion, presumably in star clusters small and numerous enough for stellar ejecta to have become uniformly mixed within the clouds. The efficiency of the early star formation must have been low to leave so much diffuse gas at z = 3; possibly the gravitational potentials of the first systems were too shallow to withstand the first stellar heating, so the mitng was accompanied by disruption.

What has become of the material in the forest clouds? Some may still be in the cool low-surface-density clouds detected in the vicinity of galaxies. Some may fuel the rapid star formation observed in less luminous galaxies at redshifls z < 1. The winds from supernovae will have redistributed some of this matter as hotter plasma. Some may have settled on to the gaseous haloes of giant galaxies, contributing to systems like that pictured in Fig. l. Some will have ended up in the plasma in groups and clusters of galaxies. One way or another, this material is now gathered together in pockets—concentrations of galaxies.

The forest clouds may be simple enough to allow analysis from first principles. Idealized models have included various geometries: spheres, slabs, filaments; states of motion ranging from quasistatic to expanding, collapsing and velocity caustics; and confinement by the pressure of hot intergalactic gas, or by the gravity of dark haloes, or simply by inertia. Recent numerical simulations suggests that the forest is indeed all of these things. Computations now have reached a level of detail that may be capable of taking account of all the relevant physical interactions, with the possibly serious exceptions of the feedback from star formation that we know is so important at the present epoch, and radiative transfer which is important in self-shielded, high-column systems optically thick to ionizing radiation. It is encouraging that the numerical simulations predict about the correct hydrogen column density distribution, coherence lengths and linewidth parameters for the absorption lines at redshifts z  ~ 2 to z  ~ 5, the largest observed for quasars. Figure 4 shows an example of the computed wispy gas distribution in the forest at z = 2.5.

In these models, structure formation commenced not much earlier than z = 5. If feedback from star formation were an important factor in slowing evolution in the forest clouds it could mean substantial cosmic evolution commenced at still higher redshift.

Summary: Cosmic Evolution Since z  = 3

Perhaps the largest difference a traveller back in time to z = 1 would notice is the effect of the general expansion of the Universe: the mean number density of giant galaxies then was an order of magnitude larger than it is now. The traveller also would see that the less massive galaxies are brighter because stars are forming faster. But if he travelled all the way back to z = 3 he would see a good deal more diffuse gas, much of it concentrated around star clusters recognizable as young galaxies, and the young galaxies would not appear strongly clustered because there is much less room for clustering than there is now.

We add some details to this picture in Fig. 5, which represents a section of space 5 Mpc by 5 Mpc on the narrow dimensions. A cube of this width at the present epoch contains on average about one giant galaxy (though there are large fluctuations around the mean). The length unit in the sketch is physical, so the mean number density of galaxies in a section this large varies as (1 + z )³ back to the epoch of galaxy assembly. The black circles represent ellipticals and the spheroid components in spirals. The evidence is that the formation of ellipticals as galaxies of stars was well under way at z = 3, and we have suggested that the similar-looking spheroid components in spirals formed early too. (To be visible the spheroids have to be drawn unrealistically large compared to the distance between galaxies, the same is true of the gaseous haloes of the galaxies.)

We indtcate gas clouds with column densities Σ_H1 >~ 10¹⁷ atoms cm^-2 by cross-hatched regions. At z <~= 1 these are found only around large galaxies. At higher redshifts they are the Lyman limit and damped Lyman-α clouds which we represent as being wrapped around young ellipticals and spheroids. Common giant spiral galaxies are represented by S-shaped curves centred on splaeroids. Following the arguments presented above, we repre sent most of the disks seen today as having originated at redshifts between z = 2 and z = 1. It is likely that there were earlier generations of disks in the damped Lyman-α clouds that were disrupted by major accretion events.

The measured small~scale correlation of the galaxy positions is consistent with statistically stable clustering at z <~ 1. Fig 5 therefore shows the nearest-neighbour distance about independent of time, and we have continued the trend back to z = 3. The result is that the voids between concentrations of galaxies are noticeably smaller at higher redshifts. The damped Lyman-α clouds at z = 3 cluster more strongly than expected from this simple picture for clustering evolution, and the young galaxies with spectra dominated by the Lyman-α emission line cluster very strongly around damped Lyman-α clouds. This may be a result of biasing—more massive protogalaxies tend to form in concentrations where the overall density is larger than average. Hence a more realistic picture might show stronger clustering of the high column density gas at z = 3. It is reasonable to suppose that the more massive spheroids that end up as elliptical galaxies tend to form in regions of higher ambient density, hence form in more strongly correlated positions, and so tend to end up in unusually dense regions, as observed.

The more widely spaced parallel lines in the sketch represent gas clouds at the column densities Σ_H1 = approximately 10¹⁴-10¹⁷ atoms cm^-2 characteristic of the forest clouds. There is still an appreciable, if much reduced, quantity of such gas at the present epoch. As we have attempted to represent in Fig. 5, this gas may be more than a megaparsec away from the nearest galaxy but it does tend to avoid the large-scale voids in the galaxy distribution. The general expansion of the Universe brings this arrangement of low-surface density gas clouds at the present epoch back to a froth that comes close to filling space at z = 3. This is a crude picture for the evolution of the forest clouds. A more detailed picture would take account of the fact that some forest material is being accreted by damped Lyman-α clouds and converted to stars in giant galaxies, some is turning into dwarf galaxies, and galactic winds and accretion shocks are redistributing some as hot plasma.

We represent dwarf galaxies in Fig. 5 by the open circles. Because the forest clouds are probably feeding star formation in dwarf galaxies, we have placed the dwarfs in or near these clouds at high redshifts. A time traveller would notice that irregular galaxies are more common at z = 1. Some of the dwarfs that contribute to the large optical counts of galaxies will by now have fallen into larger galaxies, some will have been disrupted by mass loss driven by supernovae, and others, perhaps the majority, undoubtedly have snivived as inconspicuous dwarfs.

At the present epoch the dwarf galaxtes seem to avoid the voids defined by the giants as strongly as do gas clouds. The likely explanation is that gravity has drawn together galaxies and gas to produce the present clustering of matter, in the process emptying the voids. If this is so, then any diffuse plasma at temperatures less than 10⁶ K would have been swept out ofthe voids too This is why we do not have an entry for smooth void plasma in Table 1.

Theories of Galaxy Formation

Neither our discussion nor our synthesis in Figs 3 and 5 have made use of ab initio theories for the origin of the galaxies and their large-scale distribution—the familiar “package deals” combining models for the dark matter, the nature of the primeval departures from exact homogeneity, and the background cosmology. That is in part because predictions of the theories are not yet very specific. For example, theories for the origin of the forest clouds yield about equally successful results if the total mass density parameter, baryonic plus nonbaryonic, is Ω = 1 or Ω = 0.2. The reason is that the parameters are tuned to produce the observed mass distribution on the scales of the galaxies. The simulations suggest that the Lyman-α forest forms as a natural accompaniment to galaxy formation out of primeval gas in any scenario.

A viable ab initio theory must be able to account for the observations, of course, and here are three examples of the challenges it must meet. First, the dramatic clearing of the voids between redshift z  = 3 and the present is a demanding constraint, and may be a particular problem for theories in which the total mass density in baryons plus cold dark matter is Ω  = 1. It is sometimes argued that dynamical measures (based on the relative motions of galaxies) underestimate the mean mass density because most of the mass is in intergalactic space and does not affect the relative accelerations of neighbouring galaxies referred to the general expansion. But if intergalactic space today were full of invisible haloes, why are they not seen today accreting absorption clouds as they do at high redshift. Why is intergalactic space so bare of baryon clouds and dwarf galaxies? The complementary problem if Ω  = 1 is that clusters of galaxies seem to have more than their share of baryons per dark matter compared to the predicted cosmic mean. Second, if we add enough hot dark matter to avoid unreasonably high concentrations of mass around galaxies (as in the "mixed" dark matter model), full-sized galaxies may not be assembled until z  <~ 1. This may agree with the rapid evolution of the intergalactic medium and the damped Lyman-α clouds at z  <~ 2, but not with the evidence that many giant galaxies look quite mature at z = 1, with very little building material left over in intergalactic space, and it does not agree with the properties of the damped Lyman-α clouds or what look like many well developed, if young, giant elliptical galaxies at z  = 3. Third, the computations indicate the forest clouds evolve rapidly (faster than the Rubble rate) once collapse commences. That means that models designed to assemble galaxies at high red Shift may agree with the early assembly of ellipticals but are in danger of producing and dissipating the forest clouds too early. As these examples show, the observations are leading to some sharp tests of models for the origin of the galaxies, and might even be expected to be beneficial guides to the builders of new ab initio theories.

The Hubble constants

R C Kennicutt, (Nature 381 13 June 1996 555) explained that a mature galaxy found in a distant, and therefore early phase of the Universe, is too old for its context. Is this the end of the theorists’ favourite cosmology, the Einstein-de Sitter model?

One of the goals of the Hubble Space Telescope (HST) was to find the rate of expansion of the Universe—the Hubble constant (H₀). After delays, optics problems and a succession of conflicting H₀measurements, many have wondered whether the controversy would be settled within the lifetime of HST. Better measurements of H₀ are now being made.

Reliably measuring H₀ is difficult be cause it requires an accurate distance scale far enough to permit accurate determination of cosmological expansion—at least 100 megaparsecs (300 million light years). Several secondary standard-candle distance indicators must be calbrated against a primary scale provided by Cepheid variable stars in nearby galaxies. The resulting scales have disagreed by as much as a factor of two, mainly due to a dearth of precise Cepheid distances. HST promises reliable distances for about distant galaxies.

The H₀ Key Project worked out several new distances, including that of NGC1365 in the Fornax cluster. These distances with several secondary methods gave H₀ values of 68 to 77 kms^-1 Mpc^-1. Another school using a single secondary method, the apparent brightness of type Ia supernovae, found H₀ = 55 to 61 kms^-1 Mpc^-1 . Earlier values were 43 to 53 km kms^-1 Mpc^-1. Improvements reflect better Cepheid distances obtained from HST.

Other improved secondary methods, including surface brightness fluctuation, planetary nebula luminosity function, the Tully-Fisher relation and other supernova methods, yielded H₀ between 65 and 82 kms^-1 Mpc^-1. Agreement with the primary Cepheid distances for nearby objects show that systematic errors—long the problem In this field—are being eliminated. These values are still 20 to 30% higher than the preferred values, but that is far from the factor of two disagreement that existed before.

As the uncertainty in H₀ decreases attention turns to its implications for the cosmological age problem. For a Friedmann cosmological model, in which gravity is determined by matter alone, H₀ = 70 kms^-1 Mpc^-1 implies a cosmic age af 9 to 14 Gyr, depending on the density of the Universe. For the special-case Friedmann model where the density is just enough to close the Universe—the flat Eintein-de Sitter model favoured by inflationary cosmology—the expansion age is 9 Gyr. This is flatly inconsistent with the ages of Galactic globular clusters, estimated to be about 13-17 Gyr.

If the stellar ages are correct, the Einstein-de Sitter model requires a global H₀ < kms^-1 Mpc^-1, which now seems unlikely. The age problem can be avoided, barely, if the density of the Universe is much lower—less than 20% of the closure density—or if most of it is in the form of vacuum energy, the “cosmological constant” first proposed and later abandoned by Einstein. Current observations cannot discriminate between these alternatives, but there is now hope that HST may finally fulfil the mission for which it was designed 20 years ago.

An old galaxy in a young Universe

Deep images from the HST and other instruments give us new glimpses of the Universe at high redshifts, when it was much younger. Most of the distant galaxies revealed in these images have the blue colour of young stars, various shapes, and seem to be interacting and colliding. The same surveys reveal a few red, old galaxies. A galaxy has been found at a redshift of 1.5 with an apparent age of more than 3.5 billion years.

If conventional cosmological models are correct, galaxies that old and that far away should not be there. The observation questions the Einstein-de Sitter cosmological model, and implies that at least some galaxies formed as early as a billion years after the Big Bang.

Radioastronomy has proved effective in identifying distant galaxies with a range of evolutionary properties. Galaxy 53W091 was found in a deep survey of faint radio galaxies. Imaging of 53W091 in the visible and infrared revealed extremely red colours, redder than a young galaxy at any plausible redshift and consistent with an evolved galaxy observed at high redshift. A more precise redshift, z = 1.552, came from a 5.5-hour spectrum taken with the ten metre Keck telescope. The recording of the spectrum and redshift of this object is noteworthy in itself. At 26th magnitude, this is one of the faintest galaxies ever observed spectroscopically, and to have its redshift determined using stellar absorption lines.

More distant galaxies are known. Star forming galaxies have been observed out to z > 3. What makes 53W091 unusual is the combination of redshift and age. Astronomers obtained the age by fitting the Keck spectrum to stellar population models, taking advantage of the age-sensitivity of the near-ultraviolet spectrum (1,900 to 3,300Å in the galaxy’s rest frame).

Uncertainty in the models and the effects of interstellar dust and chemical composition, which can mimic an age signature, complicate dating, but the authors bypassed the difficulties by directly fitting the absorption spectrum and the visble and infrared colours. The best age estimate is 3.5 Gyr, with a strong lower limit of about 3 Gyr. Systematic errors mainly make the age longer. This is not the first report of an intrinsically old galaxy at high redshift, but it is the strongest case, in terms of the robustness of the age determination and the high red-shift of the galaxy.

Compare this result with the expected age of the Universe at z = 1.55. The cosmic age at a given redshift depends on the Hubble constant and the density of the Universe. For a Universe with critical density (Einstein-de Sitter model) and H₀ in the range 60 to 80 km s^-l Mpc^-1 the Unverse is only 2.0 to 2.6 Gyr old at z = 1.55, over a billion years younger than 53W091. A time delay of 1 Gyr for star formation, increases the age discrepancy to 2 to 3 Gyr, and makes 53W091 roughly twice as old as an Einstein-de Sitter Universe at that redshift.

A similar problem is that values of H₀ in the range 60 to 80 km s^-1 Mpc^-1 correspond to a cosmic expansion age of 8 to 11 Gyr—far less than the 13 to 17 Gyr ages of Galactic globular clusters. The presence of luminous quasars at redshifts up to z = 4.9 (0.6 Gyr after the Big Bang for an H₀ = 70 kms^-1 Mpc^-1 Einstein-de Sitter model) may be a third age problem. Resolving these contradictions requires a low-density Universe, or one dominated by a cosmological constant.

The apparent age of 53W091 also places constraints on galaxy formation and evolution models. The galaxy must have formed at an enormously high redshift, z > 4, and the subsequent star formation must have been largely completed by z = 1.55, a span of only 1.5 to 3 Gyr, depending on the cosmology adopted.

Is the great age inferred for 53W091 a fluke? Even stronger limits may come from infrared imaging of a class of z > 3 galaxies identified using the Keck telescope. Those objects are all active star-forming galaxies, but if they possess an appreciable infrared excess from older stars they may impose even stronger constraints on the cosmic age scale, and provide further clues to the early stages of galaxy formation.

Skeptical Resources—Internet infidels | Jesus Never Existed | Steven Carr’s Website | Christianism | Early Christian Writings | God is Imaginary | “Religion Detoxification” | Our Judaio-Christian Heritage | Archaeoastronomy | No Deity | No Beliefs | Evil Bible | Bible God | ex-Christians | Jesus Police | Islamic Faith Freedom | Atheist Experience | American Atheists | Jovial Atheist | Askwhy! books
Other Resources—Early Christian Docs | Resources for Study | Traditional Bible-History | Traditional Bible World History | Traditional Bible History | about.com biblical history | Apologetics web sites | Advent Ch Fathers | Orion center links | Wikipedia | Traditional Jewish History