8.2. Alternative interpretations
Clearly, the interpretation of an accelerating cosmic expansion cannot go unchallenged (see also Riess 2000, Turner 2000). The two most serious challenges to acceleration are, in principle, evolution/progenitors and dust extinction.
The possibility that SNe Ia undergo some type of cosmic evolution cannot be easily dismissed. After all, almost everything else (e.g., the host galaxies themselves, the metallicity) evolves. The inference of acceleration, on the other hand, assumes that the local and high-redshift supernovae are drawn from the same statistical sample. I should explain that by evolution, I do not mean that somehow the processes governing supernova explosions are necessarily changing with redshift, but rather that when observing the high-redshift universe we are sampling younger galaxies, and consequently the population from which the supernovae are drawn is typically younger.
Since SNe Ia represent thermonuclear disruptions of mass-accreting white dwarfs (Livio 2001, Hillebrandt and Niemeyer 2000), changes in the C/O ratio in the white dwarfs (as a result of lower metallicities in the distant past), for example, could produce some inhomogeneity in SNe Ia light curves (Umeda et al. 1999). However, observations tend to show that even if such an effect exists, it is insignificant. In fact, existing observations of local galaxies (e.g., van den Bergh 1994, Cappellaro et al. 1997, Riess et al. 1999) already span a wider range of metallicities and host properties than that expected on the average between the sample of galaxies at z ~ 0 and z ~ 0.5.
To my knowledge, there is only one known evolutionary effect that is physically meaningful, that can mimic accelerated expansion. This is the effect of the metallicity on the density at the point of central carbon ignition (the trigger of SNe Ia). Generally, a lower metallicity (as expected at high-z) will result in a lower central density (Nomoto et al. 1997). This is because a lower metallicity results in a lower abundance of 21Ne, that is responsible for much of the neutrino cooling (via the so-called local URCA shell process, involving the 21Ne-21F pair). A lower metallicity therefore reduces the cooling and leads to an earlier ignition. Due to the lower white dwarf binding energy, the light curve exhibits a more rapid development, and a lower implied maximum brightness. However, as I noted above, the local (low-z) sample of galaxies does not appear to exhibit any significant metallicity-dependent effect.
Dust extinction is another natural candidate for making the distant supernovae appear dimmer. In fact, the observed decrease in the brightness (~ 0.25 mag) of the distant sample would require only a ~ 25% increase in the extinction. Both teams used colors to correct for dust extinction (like in sunsets, ordinary dust also reddens the light). Furthermore, observations from the optical to the infrared of SN 1999Q (at z = 0.46) showed that a large extinction to this supernova is highly improbable (Riess et al. 2000), making the dust hypothesis untenable.
Another potential uncertainty is associated with the progenitors of SNe Ia. The progenitor systems of SNe Ia are not known without doubt (Livio 2001). In particular, the two leading progenitor-system candidates are double white dwarf systems (that merge to produce the explosion; Webbink 1984, Iben and Tutukov 1984) and systems in which the white dwarf accretes from a normal companion Whelan and Iben 1973, Nomoto 1982). The possibility therefore exists, in principle, that the local and distant samples are dominated by different progenitors, with one class producing somewhat dimmer supernovae. This possibility should certainly be considered, especially in view of the fact that Li et al. (2001) find a relatively high (~ 40%) fraction of peculiar SNe Ia among the local sample.
Yungelson and Livio (2000) have shown that while it is possible, in principle, that one class of progenitors (e.g., double white dwarf systems) dominates the local sample and another (a white dwarf with a normal companion) the high-z one, such a transition is not likely, because of the following reason. If such a transition were to occur, one would expect that around the transition point (at z ~ 1), SNe Ia would be composed of an equal mix of the two classes. This is inconsistent with the observations, that show that the high-z sample is actually extremely homogeneous (Li et al. 2001). Consequently, it is highly unlikely that the interpretation of an accelerating universe is an artifact of the existence of two classes of progenitors (see Livio 2001, for a more complete discussion).
Other effects, such as those resulting from gravitational lensing, will require some further consideration once larger databases of high-redshift supernovae become available. Generally, it can be expected that most lines of sight to detected supernovae do not pass through large mass concentrations. Consequently, light paths are being bent out of the line of sight - resulting in deamplification. In rare cases however, the line of sight may cross significant matter concentrations resulting in strong amplification. The picture that emerges, therefore, is that of a brightness distribution in which the peak is shifted toward a lower value, but with a tail of bright objects (e.g., Holz 1998). Effects of this type can be averaged out once deep data from many lines of sight become available, perhaps with the proposed wide-field imager, the Super Nova/Acceleration Probe (SNAP).