The inference that our universe is dominated by dark matter is in itself a discovery of the first magnitude. But the realisation that even more mass-energy is in some still more mysterious form - dark energy latent in space itself - came as a surprise, and probably has even greater import for fundamental physics.
If this meeting had been taking place 3 years ago, the more open-minded among us would have given equal billing to two options: a hyperbolic universe with of 0.3, (in which it would be a coincidence that the Robertson-Walker curvature radius was comparable with the present Hubble radius), or a flat universe in which something other than CDM makes up the balance, equivalent to of 0.7 (In this case it would be a coincidence that two quite different invisible substances make comparable contributions).
But it is now clear that only the second option remains in the running: there is compelling evidence that the universe is flat. This evidence comes from the slight temperature-differences over the sky in the background radiation, due to density irregularities which are the precursors of cosmic structure. Theory tell us that the temperature fluctuations should be biggest on a particular length scale that is related to the distance a sound wave can travel in the early universe. The angular scale corresponding to this length depends, however, on the geometry of the universe. If dark matter and baryons were all, we wouldn't be in a flat universe - the geometry would be hyperbolic. Distant objects would look smaller than in a flat universe. In 2001-02, measurements from balloons and from Antarctica pinned down the angular scale of this `doppler peak': the results indicated `flatness' - a result now confirmed with greater precision by the WMAP satellite.
A value of 0.3 for DM would imply (were there no other energy in the universe) an angle smaller by almost a factor of 2 - definitely in conflict with observations. So what's the other 70%? It is not dark matter but something that does not cluster - some energy latent in space. The simplest form of this idea goes back to 1917 when Einstein introduced the cosmological constant, or lambda. A positive lambda can be interpreted, in the context of the ordinary Friedman equations, as a fixed positive energy density in all space. This leads to a repulsion because, according to Einstein's equation, gravity depends on pressure as well as density, and vacuum energy has such a large negative pressure - tension - that the net effect is repulsive.
Einstein's cosmological constant is just one of the options. A class of more general models is being explored (under names such as `quintessence') where the energy is time-dependent. Any form of dark energy must have negative pressure to be compatible with observations - unclustered relativistic particles, for instance, can be ruled out as candidates. The argument is straightforward: at present, dark energy dominates the universe - it amounts to around 70% of the total mass-energy. But had it been equally dominant in the past, it would have inhibited the growth of the density contrasts in cosmic structures, which occurred gravitational instability. This is because the growth timescale for gravitational instability is ~ (G c)-1/2, where c is the density of the component that participates in the clustering, whereas the expansion timescale scales as (G total)-1/2 when curvature is unimportant. If total exceeds c, the expansion is faster, so the growth is impeded. (Meszaros, 1974).
In the standard model, density contrasts in the dark matter grow by nearly 1000 since recombination. If this growth had been suppressed, the existence of present-day clusters would therefore require irregularities that were already of substantial amplitude at the recombination epoch, contrary to the evidence from CMB fluctuations. For the `dark energy' to be less dominant in the past, its density must depend on the scale factor R more slowly than the R-3 dependence of pressure-free matter - i.e. its PdV work must be negative. Cosmologists have introduced a parameter w such that p = w c2. A more detailed treatment yields the requirement that w < - 0.5. This comes from taking account of baryons and dark matter, and requiring that dark energy should not have inhibited the growth of structure so much that it destroyed the concordance between the CMB fluctuations (which measure the amplitude at recombination) and the present-day inhomogeneity. Note however that unless its value is -1 (the special case of a classical cosmological constant) w will generally be time-dependent. In principle w(t) can be pinned down by measuring the Hubble expansion rate at different redshifts
This line of argument would in itself have led to a prediction of accelerating cosmic expansion. However, as it turned out, studies of the redshift versus the apparent brightness of distant SNIa - strongly suggestive if not yet completely compelling - had already conditioned us to the belief that galaxies are indeed dispersing at an accelerating rate. As often in science, a clear picture gradually builds up, but the order in which the bits of the jigsaw fall into place is a matter of accident or contingency. CMB fluctuations alone can now pin down DM and the curvature independent of all the other measurements.
The `modern' interest in the cosmological constant stems from its interpretation as a vacuum energy. This leads to the reverse problem: Why is lambda at least 120 powers of 10 smaller than its `natural' value, even though the effective vacuum density must have been very high in order to drive inflation. If lambda is fully resurrected, it will be a posthumous `coup' for de Sitter. His model, dating from the 1920s, not only describes inflation, but would then also describe future aeons of our cosmos with increasing accuracy. Only for the 50-odd decades of logarithmic time between the end of inflation and the present would it need modification!. But of course the dark energy could have a more complicated and time-dependent nature - though it must have negative pressure, and it must not participate in gravitational clustering.