The simple model that drove us to consider an expanding universe, the linear dependence of distance on redshift, which was interpreted as a Doppler shift of the radiation, must be dropped once we start observing objects at large redshift. The reason is that the effects of general relativity creep into the observations. Two effects are particularly important: part of the redshift of the radiation is a gravitational redshift, and the angular size of the object emitting the radiation no longer follows an inverse-square law because of the curvature of space by gravity. The apparent magnitude of an object with redshift therefore becomes a test of theories of cosmological expansion.
Two groups, the High-z Supernova Search Team and the Supernova Cosmology Project, are now determining the dependence of apparent magnitude with redshift of type 1a supernovae using supernovae observed at redshifts between 0.01 and 1.2. What both groups find is that the peak apparent magnitudes of the supernovae are slightly larger, meaning that the peak brightnesses are slightly smaller, than one expects from the Friedmann cosmologies. As compared to a zero-density cosmology Friedmann cosmology, where the universe is coasting, the average peak brightness of the supernovae at z = 0.5 is lower by values ranging from 0.14±0.06 to 0.06±0.04 magnitudes, with the differences depending on the group and the analysis method. These values correspond to an an average peak brightness that is from 12% to 5% lower than expected in the zero-density cosmology.
If one accepts that these analyses are correct, meaning that there are no systematic errors in the analysis, and the type 1a supernova remains a standard candle to high redshift, then within the standard theories of expansion, one is forced to adopt a non-zero cosmological constant. The observations imply that, a given redshift, the supernovae subtend a smaller angle on the sky than is expected in the Friedmann models; this is consistent with an expansion rate for the universe that increases with time, so that the supernovae are farther way than expected in the coasting universe. A non-zero cosmological constant is a simple way of doing this.
The shortcoming of this interpretation is that it implies a peculiar coincidence: the effects of a cosmological constant are just now coming into play, and before this time, the expansion of the universe was dominated by the matter within the universe. There is no reason for this affect to appear now rather than 100 billion years from now or 10 billion years earlier. There is nothing that ties the cosmological constant to our existence, so its appearance now is simply fortuitous. For many, including the author, this provokes skepticism about the interpretation, especially given how small the effect is that is being explained.
Because the deviation from the zero-cosmological constant solutions is appearing at a redshift of order unity, an explanation of the effect tied to redshift would be more persuasive. Two that are obvious are that the redshift of the spectrum of a type 1a supernova introduces systematic errors in the analysis of the data, and the different conditions of the universe at high redshift versus at low redshift changes the luminosity produced in a type 1a supernova. Both effects are under investigation.
The analysis is difficult for several reasons. The intensity of a supernova must be well-sampled in time to determine when the peak luminosity occurs, and it must be corrected for the emission from the underlying galaxy. It must be corrected for redshift and extinction by dust, and it must be corrected for the light-curve shape. It is in these corrections that one worries about systematic errors.
A number of evolutionary effects appear at redshifts of order unity. Among these are the changes in the composition of the interstellar medium, the effects of a lower metallicity on the evolution of type Ia supernovae progenitors, and the effects of gravitational lenses.
The composition of the interstellar medium affects how interstellar dust absorbs the light from a supernova, and it affects the signatures of the presence of the interstellar medium. When dust absorbs light from a source, it reddens the source, a fact that can be used to estimate the amount of light that is absorbed. If the reddening effects change with metallicity, however, the estimates of how much light is absorbed will be wrong.
The effects of the evolution of the universe on the progenitors of type 1a supernovae is unknown, and it may be important. There are indications that the more distant supernovae are intrinsically bluer than the low redshift supernovae, which may indicate that these supernovae are dependent to some extent on the evolution of the universe.
A final important aspect of type 1a supernovae is that they do vary some in peak luminosity. Supernovae from a given redshift are an ensemble of objects that produce a variety of luminosities. If the composition of this ensemble is based on the age of the universe, then this could produce a redshift-dependent average peak luminosity. For instance, supernovae are thought to occur when a degenerate dwarf in a binary system is pushed over the Chandrasekhar limit; it take longer for this to happen in a small degenerate dwarf than a large degenerate dwarf, so the early supernovae may be weighted to stars that are initially large.
Gravitational lenses can also produce part, but apparently not all, of the observed effect. Concentrations of mass cause light that passes by to bend, as though the gravitational field of the mass were a lens. Concentrations of mass in stars, galaxies, and clusters of galaxies act as lens that deamplify the brightness of high-redshift supernovae.
A peculiarity of this scientific episode is the reaction of the scientific community; it rapidly settled on the non-zero cosmological constant as the explanation for the results. Perhaps this is a wish, because discovering new physics is more interesting than either uncovering systematic effects or refining conventional physics. My own experience in these matters, however, suggest that the more mundane possibilities are much more probable than new physics.