Martin Rees is Professor of Cosmology and Astrophysics and Master of Trinity College at the University of Cambridge.
When I enrolled as a Cambridge University graduate student in October 1964, after undergraduate work in mathematics, I had no particular research project in view, and minimal confidence that I had made the right choice -indeed I seriously thought of switching to economics. But I ended up with few regrets, because of two bits of excellent luck which I couldn't initially foresee.
First, I was assigned as one of Dennis Sciama's supervisees. I already knew of Sciama through his splendid lecture course on relativity, and had read his book The Unity of the Universe (Sciama 1959). He had charisma; he inspired his research group with his infectious enthusiasm; he followed developments in theory and observation along a broad front; and he was a fine judge of where the scientific opportunities lay. When I joined this privileged group, George Ellis had completed his PhD, and was starting a postdoc; Stephen Hawking was still a graduate student, two years ahead of me; my closest contemporaries in the group were Brandon Carter, Bill Saslaw, and John Stewart. Within a few months I felt I had made a fortunate choice.
But there was a second piece of luck: the mid-1960s were years of ferment in observational and theoretical cosmology. The discovery of the CMBR was of course the preeminent event, but these years also saw the emergence of "relativistic astrophysics:" the first high-redshift quasars, the discovery of neutron stars, and the first results from space astronomy (especially X-ray astronomy).
Dennis Sciama was "plugged in" to all these developments. He encouraged his students to interact and to learn from each other. He eagerly shared new preprints (and correspondence, news of conferences, and so forth) with his students and postdocs, and with other colleagues such as Roger Tayler. (For instance, I learned during coffee-time sessions about Hoyle and Tayler's work on helium formation, and the parallel work of Peebles. Also about the debate with the Moscow relativists about the nature of singularities.)
In the late 1940s, Fred Hoyle, Thomas Gold, and Hermann Bondi - then all in Cambridge - had proposed the steady state theory, according to which the universe, although expanding, had existed in the same state from everlasting to everlasting: as galaxies moved away from each other owing to the expansion, new atoms were continually created, and new galaxies formed in the gaps. This theory never acquired much resonance in the USA (and still less in the Soviet Union). But its three advocates were vocal and articulate people; and in the UK, especially in Cambridge, the theory was widely publicized and discussed. And it was indeed a beautiful concept. Sciama himself espoused it, and indeed described himself as its most fervent advocate apart from its three inventors.
The steady state theory was (rightly) touted as being a good theory because it was vulnerable to disproof. It made definite predictions that everything was the same, everywhere and at all times (that is, at all redshifts). Therefore if things were different in the past from now, that was evidence against it. Even if there were evolutionary changes, optical astronomers in the 1950s were unable to detect objects at sufficiently large redshifts (and look-back times) for such changes to show up. However, radio astronomers realized that some of the discrete sources detected in their surveys were "exploding galaxies" too far away to be detected optically. Although the redshifts of individual objects were unknown, it was possible to draw inferences from the relative numbers of apparently strong and apparently weak sources (since the latter would, statistically at least, be at greater distances). In particular, the number of sources brighter than flux density S would scale as the —3/2 power of S in a Euclidean universe (as discussed on page 55), and when expansion and redshift were taken into account, the log N - log S plot in a steady state universe would be flatter than the Euclidean slope. The first credible evidence against a steady state came from Martin Ryle's radio astronomy group in Cambridge (based in the Cavendish Laboratory), and from the Australian group headed by Bernie Mills. The slope (at least at the bright, high-S, end) was steeper than —3/2. Such a steep slope was incompatible with steady state. Ryle interpreted it (correctly as we now recognize) by postulating that we lived in an evolving universe where galaxies in the past (when young) were more prone to indulge in the "explosive" behavior that rendered them strong radio emitters.
For me, coming fresh to the subject in around 1964, the skepticism that greeted Ryle's evidence was perplexing. Ryle's claims - indeed everything he had claimed from 1958 onward - seemed compelling to me (and have indeed been vindicated by later developments). But I later realized that the skepticism of the "steady statesmen" was not simply irrational obstinacy. Some of Ryle's previous data, in particular the earlier 2C survey, had turned out to be unreliable, owing to "confusion" caused by inadequate angular resolution. Moreover, he had initially vehemently opposed the suggestion that the so-called "radio stars" - discrete radio sources with no obvious optical counterpart - were actually distant galaxies. To add even more irony, it was actually Thomas Gold who first made that suggestion - which of course became the cornerstone of Ryle's later argument in favor of an evolving universe. This "baggage" dating back to the early 1950s perhaps helps to explain why the steady statesmen held out against the evidence of the source counts. There was also, it has to be said, a personal antipathy between Hoyle and Ryle - two outstanding scientists of very different style.
Sciama took Ryle's data seriously, but when I joined his group in 1964 he was still clinging to the steady state theory. He conjectured that many of the unidentified sources were nearby. The apparent steepness of the log N-log S relation could then (he argued) reflect nothing more fundamental than a local deficit. But when the sources were revealed to have high redshifts, he abandoned this model (and never went along the route of saying that red-shifts were noncosmological). The clinching evidence that led Dennis Sciama to abandon the steady state was a very simple analysis that he and I did together on the redshift distribution of quasars (Sciama and Rees 1966). By 1966, more than 20 radio sources in the 3C catalog had been identified with quasars with known redshifts (extending up to z = 2.01 for 3C9). We applied to this small sample a crude version of the "luminosity/volume" or V/Vm test developed by Rowan-Robinson (1968) and by Schmidt (1968). If the universe were in a steady state, the quasars of the highest intrinsic luminosity should have been uniformly distributed in comoving volume. But when we split them into redshift bins, each bin corresponding to a shell containing the same comoving volume as the others, the quasars were concentrated in the high redshift bins. This evidence suggested that quasars were more common (or more luminous) in the past - just as Ryle had argued was the case for radio sources.
In a big bang model, the redshift distribution of quasars tells us little about the geometry of the universe, but something about the astrophysi-cal evolution of galaxies - indeed I still work on the implications of such data for galaxy formation, reionization of the intergalactic medium, and cosmic structure formation. The detection of the CMBR of course offered far stronger evidence for an evolving universe than the radio source counts. Attempts to attribute the CMBR in a steady state model to a population of discrete sources were even more contrived than those required to reconcile the theory with radio source counts and quasar data. The attraction of the steady state model was that everything of cosmic importance must be happening somewhere now, and therefore must in principle be accessible to observations. The theory's advocates believed - as was reasonable in the 1950s - that in a big bang model crucial processes would be inaccessible. But it has turned out that we can indeed observe "fossils" of the formative early eras of cosmic history soon after the big bang. The CMBR itself, of course, is one such relic; so also are cosmic helium and deuterium, and the fluctuations in the CMBR. So Sciama's disappointment was short-lived and he became quickly reconciled to the big bang - indeed he espoused it with the enthusiasm of the newly converted.
In parallel with these observation-led advances, the 1960s saw a renaissance in general relativity - a subject which had for several decades been rather sterile, and sidelined from the mainstream of physics. The impetus came from Roger Penrose. In my first year as a graduate student, I heard Penrose speak in Cambridge about his concept of a "trapped surface." I understood little of it, but was nonetheless fascinated. Roger Penrose is the kind of person who, even if you don't understand (or don't believe) what he's saying, gives the impression that an unusually insightful brain is at work. His thinking is not merely much deeper than most of us can manage - it is of a very special geometrical nature. Sciama was quick to seize on the importance of Penrose's new concepts. (Indeed it was he who had persuaded Penrose, whose PhD was in pure mathematics, to shift his interests to relativity.) Sciama encouraged some of his students to attend a lecture series that Penrose was giving in London. The most important outcome was Stephen Hawking's subsequent collaboration with Penrose, which led to the singularity theorems for gravitational collapse. The main import of Penrose's work for cosmology - as described in the article by George Ellis on page 379 - was an adaptation of these arguments to show that there must have been a "singularity" in the past of our universe, even if it was irregular at early times.
There was, at that time, a substantial research effort (spearheaded by George Ellis and a series of collaborators) aimed at investigating and classifying the various classes of homogeneous but anisotropic cosmological models. This was an interesting exercise in its own right. However, a special motivation came from Charlie Misner, who spent the academic year 1966-1967 on sabbatical in Cambridge. It was from Misner that we learned about the so-called "horizon problem," that causal contact becomes worse in the early phases of a Friedmann (decelerating) universe, rendering it a mystery that the present universe seemed so uniform and synchronized. Misner noted that causal contact would have been better if the early expansion had been anisotropic - best of all in the "mixmaster" model where there was an alternation in the axes of fast and slow expansion. The aim of the "Misner program" was to show that a universe could have started off (and homogenized) via a mixmaster phase, but that the initial anisotropies would have been erased, either dynamically or via neutrino viscosity. This program failed - and until the invention of the "inflationary" universe, more than a decade later, most of us probably thought that an explanation of global homogeneity would have to await a quantum-level understanding of the singularity. It was coincidental that the theoretical advances in relativity, instigated by the new "global methods" that Penrose pioneered, happened concurrently with the discovery of the CMBR.
It was a further coincidence that, during the 1960s, objects were discovered where general relativity was crucial, rather than a trivial refinement of Newtonian gravity - discoveries that stimulated the new research area of "relativistic astrophysics." The discovery of quasars (and, later, of neutron stars) indicated that objects probably existed in which the crucial features of Einstein's theory would have to be taken into account. Black holes of course are the most remarkable prediction of Einstein's theory. The Schwarzschild solution, discovered in 1916, represents the simplest black hole. They were speculated about in a rather half-hearted way by astronomers and cosmologists in the 1930s to 1950s. But the term "black hole" was not used until 1968, when it was coined by John Wheeler, and it was only in the late 1960s that theorists really clarified the nature of black holes.
A more general solution, discovered in 1963 by Roy Kerr (1963), was believed to be a description of a collapsed spinning object. The biggest breakthrough actually came from the work of Israel, Carter, Hawking, and others. They showed that Kerr's solution was generic, in the sense that any black hole would end up being described by this particular solution of Einstein's equations. Any gravitational collapse leads, after the emission of gravitational waves, to a black hole described exactly by two numbers, its mass and its spin. So black holes proved to be just as standardized as an elementary particle.
The number of people involved in these theoretical developments was even smaller than the experimental and observational community - indeed most relativists were associated with one of three "schools," those centered in Princeton, Cambridge, and Moscow. Communications were far less immediate than today (especially, of course, between East and West in the Cold War era). However the interactions that occurred were almost invariably cooperative and friendly. My own work was mainly on astrophysics and on galaxy formation: for this work, the new paradigm of the hot big bang was the essential backdrop, rather than being at the focus. My aim was to understand how galaxies produced so much radio power, how they became quasars, etc. It was already fairly clear that the power generation involved gravity, although, despite early advocacy by Salpeter, Zel'dovich, and Novikov and (especially) Lynden-Bell, it wasn't as clear as it would become in the 1970s that a single huge black hole was implicated.
I continued to be uneasy, until the early 1970s, about the apparent coincidence between the energy in the CMBR and the energy that could be supplied by astrophysical sources (via hydrogen burning or via gravitational collapse), but this proved of course a blind alley and distraction.
I can lay claim to two minor positive contributions directly related to the CMBR. One (Rees and Sciama 1968) concerned what is now sometimes called the "Rees-Sciama effect" - the perturbation in the CMBR due to a transparent gravitational potential well along the line of sight (for example, a cluster or supercluster of galaxies). In the linear regime, this is subsumed in what is normally called the "integrated Sachs-Wolfe effect" - it is nonzero except (to first order) in the Einstein-de Sitter universe. However there is a distinctive effect due to virialized clusters. Had Sciama and I known then the actual amplitude and scale of clustering, we would not have felt it worthwhile to explore these higher-order effects. But at that time there was no way of ruling out large-amplitude density fluctuations on gigaparsec scales (indeed there were early - and in retrospect misleading - indications of such clustering from the distribution of quasars over the sky). This effect has only recently been detected. My second contribution (Rees 1968) addressed the possible polarization of the CMBR. The simplest illustrative examples of this effect arose in anisotropic but homogeneous models (though the effect was obviously present in more general models). This work stimulated an early search by Nanos (1974, 1979), but it was more than 35 years before polarization was actually detected.
In CMBR studies, a consensus has generally quickly developed whenever there has been an advance - this is in contrast to (for instance) the prolonged debate and perplexity about the physics of AGNs and quasars. This is because the CMBR data, though challenging to obtain, are "cleaner," and the relevant fluctuations are in the linear regime. Successive developments - the CDM paradigm, the CMBR fluctuation spectrum, and so forth -have led to a well-established set of cosmological parameters. It has been a privilege to have followed a subject where progress has been sustained so consistently for 40 years, and to have known many of the scientists to whom these historic advances are owed.
Was this article helpful?