intensity profile of the universal CMBR—a curious connection between the smallest and the largest realmsof physical theory. The simplest and most compelling explanationfor the thermal shape of the CMBR is that the universe was all atthe same temperature for some very early part of its early history.Space was uniformly filled with hot (but rapidly expanding and cooling)ionized gas and thermal radiation. The recognition that such an epochexisted in the early history of the universe is a cornerstone ofmodern cosmological models.
That a single experiment can have such profound implications is unusual.But the measurement was not easy. It required the use of a satelliteto get the instrument above the atmosphere of Earth, and the entireinstrument was cooled to a temperature of 1.5 K to reduce radiationfrom the instrument itself. The design, operation, and calibrationof the COBE instrument all had their technological heritage in ground-basedand balloon-based instruments, and the experience gained in theseearlier experiments laid the foundation for the highly successfulsatellite measurement.
Theoretical modeling of the thermal history of the universe has developedconcurrently with progress in the spectral measurements. The accuratefit of the measured CMBR spectrum to a thermal shape sets extremelytight bounds that limit the variety of hypothetical physical processesthat could have taken place in the early universe. For example, energy-releasingprocesses that would have reheated the universe at critical epochsin its history can now be ruled out.
Cosmological theories based on the Big Bang unambiguously predictthat the temperature of the universe will fall with time as the universeexpands. Recent observations from the new Keck telescope have providedthe first direct evidence that the CMBR temperature has indeed decreasedover (relatively recent) cosmic time. Using the light from a distant,bright, quasi-stellar source, astronomers were able to measure thetemperature of carbon atoms in an intergalactic cloud between Earthand the source. The shift of the spectral lines (redshift) of thecloud was also measured. The temperature of the cloud, which shouldhave been the same as that of the CMBR, was found to be 7.6 K. Thisis exactly the value expected for the CMBR at that redshift.
Why are “bumps” in the CMBR so important?
The second important characteristic of the CMBR is the variationin intensity (or temperature) from place to place on the sky. Measurementsof these variations, often called anisotropy measurements, tell usabout tiny fluctuations in the uniformity of the early universe.Though small (1 part in 100,000), these fluctuations are believedto be the seeds of all complex structure in the universe today.
The main idea is that the universe as we see it today must somehowhave evolved from the early stage of thermal uniformity implied bythe CMBR spectrum measurements. Today, the universe is a rather livelyplace with everything from stars and planets to quasars, collidinggalaxies, and black holes. A central unresolved question is, Howdid these structures form? One of the most appealing answers is alsothe simplest: objects formed because gravity pulled together matterthat existed originally in slightly denser regions. Given enoughtime, the matter became compressed because, as more matter was pulledin, the gravitational forces grew even stronger until galaxies andother objects resulted. These objects resist further collapse becauseof their rotation and/or internal motions.
This idea, called gravitational instability, probably explains mostof the objects that astronomers see in the sky today, but it requiressmall initial density fluctuations to start the process of collapse.Because the expansion of the universe greatly retards the formationof instabilities, the seeds of structure must already have existedat the time the CMBR last interacted with matter, 150,000 years afterthe Big Bang. Fluctuations that existed then are detectable by theCMBR anisotropy measurements as tiny variations (bumps) in radiationintensity across the sky.
Measurements of Anisotropy
The theoretical details of the fluctuations and their consequenceshave been developed in the 30 years that cosmologists have knownabout the CMBR. Until 1992, no anisotropy (except for a separateeffect due to the motion of our Solar System in the cosmos) had beendetected in the CMBR, though many attempts had been made. Increasedreceiver sensitivity forced experimenters to develop increasinglysophisticated techniques to reduce systematic errors and the effectsof noise from other sources of microwave radiation. As measurementsbecame more sensitive and no anisotropy was found, the range of theoreticalmodels that could fit the observations became smaller and smaller.As models were increasingly constrained, many cosmologists, especiallythe theorists making predictions, became increasingly nervous overthe lack of detected fluctuations. So it was with great excitementthat the COBE science team announced in 1992 that it had detectedthe long-sought bumps in the CMBR. The illustration on the coverof this report is the resulting COBE map of the intensity of microwaveradiation arriving from various directions in the sky. The map containssome instrumental noise, but its lumps and bumps also show evidencefor the beginnings of structure in the universe.
Like the measurement of the CMBR spectrum, the COBE detection ofthe CMBR anisotropy could not have occurred without the experienceof earlier anisotropy experiments from the ground, from balloons,and from aircraft. As valuable experience was gained from the suborbitalmeasurements, the technology and the experimenters' understandingof how to avoid contamination from many bright local sources bothevolved. Balloons, rockets, and aircraft also provide important opportunitiesto follow up on satellite discoveries. Recently, data from an independentballoon experiment, using a frequency above that of the COBE's receivers,exhibited the same basic CMBR pattern as seen in the COBE data, thusconfirming the satellite result. Lowcost balloon experiments havealso enabled important first steps toward extending these resultsto smaller angular scales.
The detection of large-scale anisotropy has finally allowed the fieldof CMBR research to become established. Whereas previous noise-limitedmeasurements could exclude but not support certain theories, theanisotropy measured by the COBE satellite is approximately at thelevel needed for the origin of structure as predicted by theories.In addition, the manner in which the strength of the anisotropy varieswith the angular size of the bumps is consistent with Big Bang theory.Even the idea of an inflationary epoch in the early universe (seesection V) seems to fit with the COBE result, although this characteristicof the CMBR fluctuations is not yet well determined. Within the pastyear, both theoretical thinking and experiment planning have undergonean important transformation because of the COBE detection. Almostall areas of cosmology have been affected, and many now take themagnitude of the COBE anisotropy signal as a reference point fornew developments.
While the COBE anisotropy detection is extremely important, it wasmade at angular scales of more than 10 degrees on the sky, scalesmuch larger than those actually involved in the formation of galaxiesand clusters of galaxies. The best direct comparison between theprimordial seeds and present day structures awaits the reliable detectionand detailed mapping of CMBR anisotropy on smaller angular scales,comparable to the physical scale of superclusters of galaxies.
Medium-scale (0.5- to 10-degree) anisotropy measurements also probeimportant details of the decoupling process. Numerical calculationsshow that during the time that matter was combining into atoms andinteracting with the CMBR for the last time, there were acousticoscillations in the overdense bumps that should have left strong
“fingerprints” in the CMBR anisotropy. Oscillations would amplify bumps to an extentthat depended on the bump size and critical details of the cosmologicalmodel. The theory's prediction of peaks in the CMBR medium-scaleanisotropy offers an intriguing and unexpected opportunity. If theprofile of CMBR anisotropy versus bump size can be accurately measured,cosmologists may be able to measure three important cosmologicalparameters: Ω, the ratio of the mean mass density to that requiredto close the universe and eventually stop its expansion; ΩB, the contribution to Ω from ordinary (baryonic) matter; and the expansion rate of the universe(the Hubble constant, H0). Of course, the data may reveal something completely unexpected,a clue to unknown processes in the early universe. In either case,medium-scale anisotropy signals are bringing us detailed informationabout the conditions and dynamics that existed in the universe whenit was only 150,000 years—old something quite unimaginable only afew years ago.
The opportunity to gain such important knowledge has recently focuseda great deal of experimental and theoretical research on the issueof CMBR anisotropy on angular scales around 1 degree, at the lowend of the medium scale. This scale corresponds to about 300 millionlight-years, which is smaller than the size of the region in theuniverse around us that has been well mapped in redshift surveysof thousands of galaxies. Experiments on this angular scale are relativelynew compared to studies of the CMBR spectrum and the large-scaleanisotropy measurements. The techniques and technology needed toovercome the new experimental challenges are just now beginning tobe understood.
A current problem is to remove possible interference from weak radiosources and faint, diffuse emission from our own galaxy. To effectthe separation, a wider range of frequencies and better sensitivityare being used. At angular scales around 1 degree, the greatest understandingof the structure-formation process would follow from detailed mappingof large regions of the sky; this poses difficult challenges forcurrent experiments. Groundbased experiments must overcome fluctuationsin atmospheric emission that are thousands of times larger than theexpected CMBR anisotropy signal. Thus, experimenters are observingfrom sites with a cold, dry atmosphere like the South Pole, northernCanada, and mountain tops. Balloon experiments afford much smalleratmospheric fluctuations, but they are limited by the relativelyshort times available for observations. Long-duration balloon flightsof many days are being planned to alleviate this problem, and theyoffer a powerful new opportunity to extend COBE results at relativelylow cost. However, flight opportunities for long-duration balloonsare currently scarce. Limited sky coverage and Earth's radiationenvironment are also problems for experimenters trying to map theCMBR anisotropy from ground-based and balloonbased platforms.
The situation for the measurement of medium-scale anisotropy is similarto that for measurement of the spectrum and large-scale anisotropytwo decades ago. Data are beginning to indicate that something interestingis happening, but greater accuracy and broader sky coverage are neededto extract the important scientific results. Techniques for successfulexperiments are being developed using the experience gained withsuborbital experiments, but Earth's environment poses major problemsfor experiments requiring high accuracy and large sky coverage. Researchnow under way will ultimately lead to the design of a satellite thatcan utilize the advantages of the space environment. Preliminaryfeasibility studies indicate that a midsize Explorer satellite inan orbit far from Earth is an attractive and relatively inexpensivepossibility. A satellite experiment will permit the mapping of theCMBR sky pattern with sufficient detail and sensitivity to form anexcellent picture of the early development of structure in the universe.
Measurements of the CMBR anisotropy on angular scales smaller thanabout 0.5 degree
must be made with relatively large groundbased telescopes (high angularresolution requires large antennas). The theoretical case is notyet as strong for small-scale measurements because such anisotropiesare presumed to have been smeared out when matter and radiation interactedfor the last time at the epoch of photon decoupling. However, variantsof the Big Bang theory, such as theories involving cosmic strings(discontinuities in the structure of space), predict that importantclues to the universe's history might be embedded in the CMBR atthese angular scales. A few experiments have been done from singlelarge telescopes and from arrays of radio telescopes, such as theVery Large Array. Though sensitivities are comparable to the COBEdetection level, only a small fraction of the sky has been scanned.Signals are detected, but they are thought to be due mostly to radioemission from galaxies, quasars, or other foreground radio sources.Because of the need for large instruments and the extreme care requiredfor these measurements, progress on small-scale anisotropy is likelyto be relatively slow. But it should be remembered that only a decadeago there was almost no interest in even medium-scale measurements.As understanding has grown, so also has the need for more diverseexperimental data. And surely, as more data are analyzed, the simplemodels of structure formation must break down at some point. Improvementsin small-scale measurements are one way to find such weaknesses inthe models.
III. THE LARGE-SCALE STRUCTURE OF THE UNIVERSE
Galaxy Maps and Large-Scale Structure
What is large-scale structure, and why is it important?
Although the intensity of the CMBR is extremely uniform in all directions,with fluctuations measured at only 1 part in 105, the local distribution of galaxies is extremely irregular, withfluctuations in the density of galaxies per volume of space beingwell in excess of 100 percent. Maps of the distribution of galaxiesin space reveal a remarkable pattern of thin, filamentary structuresconnecting small and large central concentrations of galaxies, punctuatedby large, quasi-spherical voids. The example of the map shown inFigure 2 (p. 4) is the result of several years of painstaking spectroscopicobservations with modest-size optical telescopes. The far-flung distributionof galaxies in the universe, the complex assemblage of clusters,filaments, and voids, is referred to as large-scale structure.
It is not surprising that galaxies are clustered. As explained above,the early universe contained small density irregularities, as measuredby fluctuations in the CMBR, and the amplitude of these small bumpsgrew via their self-gravity to make the structure seen today. Thiscondition of gravitational instability can amplify the initial densityfluctuations of seeds on all scales. Galaxies and large-scale structureare all part of the same process; both are relics of the Big Bang.
Finding clumps of galaxies was thus expected, but their huge extentcaught astronomers by surprise. Typical voids are 200 million light-yearsacross, and one enormous curtain-like structure—the Great Wall—is draped acrossthe universe in a span half a billion light-years across. Even thislarge size, however, is less than a tenth the scale measured by theCOBE satellite discussed above. Altogether, the distances involvedin the study of large-scale structure range over a factor of a million,from the size of galaxies to the CMBR anisotropy measured by theCOBE satellite. This combination of observations gives us a powerfulprobe of Big Bang density. fluctuations over a wide range of scales.The extent of early density fluctuations on different size scalesand their subsequent growth under gravity are critical clues to thenature and amount of dark matter in the universe, as explained below.
Mapping the large-scale structure
Making maps of galaxies in three dimensions requires knowing howfar away each galaxy is from Earth. One way to get this distanceis to use Hubble's law for the expansion of the universe. Hubblediscovered that the velocity at which two galaxies recede from eachother is proportional to the distance between them. Inverting thisrelation yields an estimate of distance from observed velocity. Thevelocity with which a galaxy is receding from us is obtained by measuringthe shift to redder colors of spectral features in its spectrum,a “redshift” analogous to the familiar Doppler shift in the frequency of soundwaves from a receding source. The greater the redshift, the largerthe velocity, and, by Hubble's law, the larger the distance.
We are truly living in the age of mapping the universe. The lastdecade has seen a revolution in the technology of light detectorsthat has made it possible to measure redshifts rapidly, even withmodest-size telescopes. In 1976, there were only 2,700 galaxies withmeasured redshifts—now there are 100,000. By the year 2000 astronomersexpect 1 million! This field of astronomy is still on a steep discoverycurve.
The importance of uniform galaxy surveys
The first step in making a redshift survey is compiling a catalogof galaxy positions and brightnesses on the sky. Traditionally suchcatalogs have been based on photographic surveys taken in visiblelight. We are learning, though, that even small biases in the listof target galaxies may have a big effect on the final maps. Hencethere is strong interest in new and
better ways of finding galaxies. Three basic avenues are being explored.Deeper surveys of the whole sky in visible light are being conductedusing highly sensitive detectors (called charge-coupled devices;CCDs) that can detect intrinsically faint galaxies, galaxies of lowsurface brightness, and distant galaxies. Near-infrared surveys ofthe sky at 2-micron wavelength will make it possible for the firsttime to observe the dip down near the plane of our galaxy, whosedust clouds obscure 35 percent of the sky in visible light. Finally,x-ray satellite surveys provide yet another means of mapping clustersof galaxies.
A fundamental question is whether one large section of the universelooks like another. In other words, how large do sections have tobe before they begin to appear statistically uniform? Unfortunately,a single ground-based observatory sees only a portion of the sky.To achieve the high degree of uniformity needed over the whole skyrequires careful surveys. There are three requirements: First, individualsurveys must cover as much of the celestial sphere as possible. Second,surveys must be closely coordinated and well standardized so thatthey can be knitted together. Finally, homogeneous all-sky surveysneed to be conducted by satellites above Earth, such as the InfraredAstronomy Satellite (IRAS), a joint mission of the United States,the United Kingdom, and The Netherlands that was flown in 1983.
A major goal for the next generation of surveys is to increase theirrange out to 3 billion light-years, roughly 20 percent of the radiusof the visible universe. On such scales cosmologists would be probingstructures that are the same size as the smallest structures in theCOBE microwave map. The clustering behavior of galaxies over an extremelywide range of scales could be measured and compared directly to theCMBR anisotropy with no extrapolation. This would tell us how thedensity fluctuations have evolved from the epoch of CMBR emission(the epoch of photon decoupling) to the present. This informationwould yield essential clues to the amount and nature of dark matterin the universe.
Theory of large-scale structure
Statistical description and theoretical modeling of the observedgalaxy distribution have been extremely productive over the pastdecade. Much of this modeling has been done with large computer simulationson the largest available supercomputers. This is a problem in the“grand challenge” class, with the goal of understanding in detail the formation ofstructure on both small and large scales. The models typically followthe evolution of a large patch of the universe. Calculations startwith random initial fluctuations as statistically predicted for differentcosmological parameters and different types of dark matter. The equationsgoverning the gravitational coupling, as well as other physical processes,are then solved numerically by the computer. Starting from smallamplitudes, the fluctuations become increasingly larger, as expectedfrom the gravitational instability picture. The computational resultscan then be compared to the observed properties of large-scale structurein the universe. With careful analysis, such comparisons can setconstraints on the amount and nature of dark matter. Some proposeddark-matter candidates have already been ruled out in this way. Figure 3 is a recent example of a numerical simulation, processed with similarselection criteria as for observational redshift surveys. The similarityin the voids and filaments shown in Figure 2 and Figure 3 is striking.
Clusters of galaxies, with size on the order of 3 million light-yearsand mass of 1015 Suns, are central to our understanding of structure. Astronomershave recently discovered that galactic cores are dense enough toact as gravitational lenses (discussed in section IV in “Gravitational Lenses”), that most of the baryonic (ordinary) matter within them is inthe form of hot gas, not galaxies, that dark matter constitutes approximately80 percent of their total mass, and that they show a considerableamount of substructure when examined at high spatial resolution.The abundance of clusters